Founder & CEO, Valiotti Analytics | Better Data Analytics for Data-Driven Companies

Founders don’t lack data.
They lack confidence to make decisions with it.

I work as a Fractional Chief Data Officer for scaling tech companies, when analytics becomes business-critical — but hiring a full-time CDO is too early or too risky.

I’m the founder of Valiotti Analytics, a data consultancy trusted by 50+ teams globally. I work with companies across the US and Europe.

I have a PhD in data science, which helps when problems are complex — but my work is always grounded in business outcomes.

What I typically change:
• Data & analytics strategy aligned with business goals
• Clear metrics and decision frameworks
• Analytics org design (people, tools, processes)
• Systems that teams actually use, not just maintain

Before consulting, I was an in-house operator. I’ve lived through messy data, slow reports, and dashboards nobody trusts — and I know why “just build another report” never works.

Since 2019, I’ve helped marketing, sales, and product teams:
→ Scale revenue up to 25x
→ Eliminate manual reporting and data chaos
→ Turn analytics into a decision-making system, not a cost center

I usually work in a fractional format (3–12 months), directly with CEOs, CPOs, and CMOs — with optional execution support from my team.

Nikolay Valiotti

The hill I’ll die on:
“Natural language → SQL” will never — I said NEVER — replace analysts.

Every few months someone announces the same miracle:
“We cracked it — business users can query data in plain English now!”

And every time I think:
Great. We solved the part of analytics that was never the problem.
SQL was never the bottleneck, and typing isn’t the job, thinking is.

Most companies don’t struggle because nobody knows how to write a JOIN.
They struggle because nobody knows what question actually matters, or how to connect it to the business reality they’re living in.

Give someone a “natural language → SQL” tool and here’s what happens:
→ They ask vague questions.
→ The model confidently answers the wrong thing.
→ And everyone walks away thinking the data is broken.

The truth is that if analytics fails, it fails long before the query — at the decision, the framing, the context, AKA the part AI can’t do for you.

This tech won’t replace analysts. It just exposes how much of analytics depends on real judgment, not syntax. If anything, AI is about to make analysts more valuable, because someone still has to think clearly.

→ Follow Nick for the data takes that won’t waste your time.
♻️ Repost if you’re tired of seeing “natural language SQL” demos pretending to solve a thinking problem with a typing feature.

Nick Valiotti

Fractional CDO | Helping Scaling Tech founders turn data into faster decisions | Founder @ Valiotti Analytics

Hard pill to swallow:

Excel was never the problem.
Dependence on Excel was.

Excel is brilliant.
It’s flexible, familiar, and absurdly powerful for its size.
But here’s the trap: the better Excel works, the more you rely on it. And the more you rely on it, the harder it is to grow beyond it.

Every company reaches that moment:
→ one file becomes 12
→ 12 becomes 40
→ 40 becomes a folder called “final_v7_really_final.xlsx”
and suddenly nobody knows which version is the truth

This isn’t a tooling problem. This is an infrastructure debt problem. And it compounds silently.

In my book “Your Fractional CDO” (Amazon: https://lnkd.in/dWqxidfk) I break down exactly why spreadsheets become dangerous at scale — and what to do before your dashboards, KPIs, and decisions start contradicting each other.

Because the solution isn’t to “ditch Excel.” The solution is to stop using it as your entire data strategy.

Here’s where companies should shift their focus:

1. Centralize your sources. Stop exporting CSVs like you’re throwing confetti. Put your data in one place — Sheets, a small warehouse, whatever. One source of truth beats ten perfect reports.

2. Assign ownership. If everyone is responsible for the data, no one is responsible for the data. Quality only scales when accountability does.

3. Automate the pipes, not the panic. Connect tools directly. Even Excel works beautifully as the front end if the plumbing behind it isn’t a disaster.

4. Measure less, decide more. 50 KPIs don’t make you data-driven. A small set of metrics that the business actually uses — does. Excel absolutely has a place in a modern stack. Just not as the backbone of an entire company.

Tools don’t create clarity. Frameworks do. Excel should absolutely stay in your stack — just not at the center of it.

So fix foundations before foundations fail you.
And get your copy of Your Fractional CDO — unless you truly enjoy living dangerously with formulas that break when a butterfly sneezes.

Most dashboards show data.

This one supports decisions. 

From a business lens, it works because it ties volume, monetization, and engagement together in one place. A leader can quickly understand:

→ Where demand sits 
→ Where revenue concentrates 
→ Where content investment goes 
→ Whether engagement backs up growth

The subject breakdown isn’t decorative. It helps with portfolio focus. 
The pricing view supports monetization strategy. 
Students vs Reviews and Students vs Lectures help justify where to invest next. 
Paid vs Free anchors the revenue model conversation.

Nothing here feels academic. It’s built for operators who need to decide where to double down — not admire charts.

Great job, Maureen Okonkwo!

If you want to build dashboards people actually use, start before you open Tableau or whatever tool you’re using. Grab my free Dashboard Canvas interview template — it helps you design around decisions, not visuals: https://lnkd.in/dcJGqeWj

undefined undefined

ClickHouse Developer

Tableau Desktop Certified Associate

A Course for Executives

Digital Transformation Academy

Founder @ Valiotti Analytics | Marketing & Sales Dashboards for Growth-Ready Companies | PhD & MA in Data Analytics

🚀 Founded valiotti.com — the go-to data agency for fixing messy marketing analytics.
Top-rated on Upwork & Clutch. Trusted by growth-stage teams to build dashboards, clean up data chaos, and drive faster decisions. Partnered with AWS, Google Cloud, and Salesforce.

Ex-Head of Consumer Marketing at the largest eastern retail company.
I’ve been the one chasing targets with thousand-row tables, slow reports, and siloed data — I know the pain.

Since 2019, I’ve helped marketing and sales teams:
→ Scale revenue up to 25x
→ Save hundreds of hours on reporting
→ Get insights that actually drive action

Let’s talk: https://calendly.com/nick-valiotti

Let's discuss your project and grow you business with the power of data

Join Friday Data Snack for weekly data, leadership, and business insight

Tableau Ambassador || Tableau Featured Author 2024 || 4x #VOTD || Tableau Most Notable Newbie Winner || Tableau Community Highlight || Business Inteligence Analyst || Data Visualization Developer

A lot of funnel dashboards are just progress reports.

This one by Varun J. is a control panel.

From a business perspective, what makes this strong is that it separates volume from efficiency, and shows both at the same time.

You can see how many users enter the system.
You can see how many convert.
And you can see how long it takes.

That matters because growth problems are rarely just “not enough traffic.” Sometimes it’s friction. Sometimes it’s delay. Sometimes it’s a specific step bleeding value.

The time-to-account-opening section is especially smart. Median time, percentiles, distribution — that’s operational insight. A leader can immediately ask:

→ Are we losing people because the process is slow?
→ Is the long tail hurting us?
→ Is speed a competitive advantage?

The time-limited conversion blocks are even more practical. Conversions within 5 minutes vs 10 minutes? That’s product design territory. That’s where small UX changes can move revenue.

And the drill-down by state and age group turns this from a generic funnel into a segmentation tool. Now it’s not just “conversion dropped.” It’s: where and for whom?
That’s how you run experiments with intent instead of guessing.

This dashboard doesn’t just describe the onboarding journey.
It exposes where to intervene.

If you’re building funnel dashboards, start with the questions the business needs answered before you start placing charts.

Data and BI @ Ledger Analytics Engineer and Data Product Developer | 2xTableau Ambassador | IITK ’21 | ESSEC MiM

The secret to being data-driven?
It’s not a secret.

It’s 3 steps and zero BS.

Step 1: ETL (aka “just move the damn data”)
Fivetran — Free up to 500K rows. Unless you’re Google, you’re fine.
Python script — For people who are either clever… or broke.
Goal: one place for your data. Not duct tape. Not Excel exports.

Step 2: Data Storage
BigQuery. That’s it.
Cheap. Scales forever. Doesn’t cry when you throw SQL at it.
Storage = pocket change.
Queries = fine unless you go full goblin mode.

Step 3: Dashboards That Don’t Suck
Looker Studio. Tableau. Whatever.
A good analyst makes any BI tool sing.
Make it visual. Make it clear. Make it clickable.
That’s how you get decisions, not headaches.

Bonus: Real-Time?
Use Segment if you want click-by-click product tracking.
Just don’t stream straight into BigQuery unless you enjoy surprise CFO meltdowns.

That’s the starter pack.

No enterprise fluff. No shiny $10K toys.
Just a stack that works.

----------

♻️ Repost if you’re still drowning in CSVs with a dream.
Follow Nick Valiotti, for stacks that don’t suck.

Quick reality check:
If your data analyst feels like an IT intern, you’re actively destroying one of the most strategic roles in your business.

Analysts aren’t “report people.” They aren’t back-office technicians. They aren’t support staff for whatever broke today.

They are business operators.
And if they’re not sitting with marketing, product, and ops, you’re wasting their entire potential.

When analytics lives under IT, this is what you get:
→ Numbers with no story
→ Dashboards with no decisions
→ “Insights” nobody actually uses

Put an analyst in the room where decisions are made, and everything changes:

Marketing: “We’re launching a new campaign.”
Analyst: “Here’s what worked before, what failed, and where your budget will quietly bleed if you’re not careful.”

That’s the job. Not maintenance. Not admin.
IMPACT. 

If your analysts don’t understand the business, they won’t produce value.
And that’s a leadership failure — not an analytics one.

♻️ Repost if you’ve seen analytics trapped where they don’t belong.
→ Follow Nick Valiotti for hard truths on building data teams that actually move revenue.

Hot take:

Data doesn't create value.
Decisions do. Data just makes decisions less wrong.

Most companies I audit have dashboards, a warehouse, analysts, sometimes a full data team. And every meaningful decision still gets made on gut feeling in a boardroom.

The chain isn't broken. It's snapping at the same point every time.
Data → Insight → Action → Measurement

Here's where each link fails:
1. Data → Insight 
Nobody's asking the right questions.  Analysts spend 80% of their time cleaning, 20% analyzing. Stakeholders request "one more cut" of the same report. 
Dashboards show what happened. Never why. Never what to do.

2. Insight → Action 
This is where most companies die. The weekly analytics review happens. Everyone nods. Someone says "interesting." The Slack message with the chart gets read once and forgotten. Nothing changes. This isn't a data problem, it's an org problem.

3. Action → Measurement 
You made the decision. Did it work? Nobody checked. Six months later the same problem resurfaces in a different dashboard. There's no loop back.

The fix for all three is the same: every analysis ends with a specific recommendation, a named owner, and a deadline.

Not "churn is up 15%."
But "churn is up 15% in enterprise due to onboarding failures, restructure onboarding for accounts >$50K, owner: Head of CS, deadline: 2 weeks."

If your data team can't name 3 decisions they influenced this quarter, that’s a decision-making problem, not a data one.

💬 Where does your chain break?
📘 If this hit close to home, I wrote a whole book for exactly this situation. "Your Fractional CDO" is for leaders who know they have a data problem but aren't sure where to start → https://lnkd.in/gaC23sz2

On International Women’s Day, I wanted to highlight a few women in data whose work I always enjoy seeing in my feed.

People who consistently share thoughtful ideas, useful insights, and perspectives that push the field forward.

Morgan Depenbusch, PhD — great examples of how design and narrative turn analysis into something people actually understand.
Clare Kitching — sharp, grounded perspectives on AI and how organizations actually adopt it.
Charlotte Ledoux — one of the voices reminding the industry that governance isn’t bureaucracy, it’s what makes trusted data possible.
Kat Greenbrook — constantly showing how insight only matters if the story behind it is clear and actionable.
Anastasiya Kuznetsova — always clear data visualizations that make complex information accessible.
Mariska van Kan — consistently strong dataviz work and perspectives on visual communication.
Donabel Santos — an incredible voice supporting people building careers in data.
Dawn Choo — thoughtful content on the intersection of data science and career growth.
Hannah Rounds — practical insights from the day-to-day world of data analytics.
Alex Wang — smart perspectives on the evolving AI & data science landscape.

One of the best things about the data community is how much knowledge people openly share. These are just a few voices I’m always happy to learn from.

If there are women in data whose work you enjoy, feel free to add them below so more people can discover them.

Happy International Women’s Day 👩‍💻 👩‍🎓 👩‍🏫 🌿

HR Data Storytelling & Influence → Turn people data into recommendations leaders trust • Corporate trainer & Keynote speaker • Snowflake, ex-Google

Transform your AI & data ambition into action | xQuantumBlack, xMcKinsey | Global top 100 Innovators in Data & Analytics | AI & data strategy, governance and capability building

Data & AI Governance Expert 🔎 | Entrepreneur | Author of The Data Governance Playbook | Speaker | LinkedIn™️ Top Voice in AI 🇫🇷

Author of The Data Storyteller's Handbook 📘 Founder of Rogue Penguin 🐧

BI Lead | Tableau Ambassador | DataViz Consultant

Dashboard product designer • I audit, rescue and educate dashboard products • Good Design Award Winner • Power BI & UX

Empowering Data Professionals Through Education | Teacher, Data Leader, Author, YouTube Educator | teachdatawithai.substack.com

Data Scientist (ex-Meta, ex-Amazon)

Data, Analytics & AI Leader | Data Governance | Manufacturing & CPG | Get Value From Your Data

Learn AI Together - I share my learning journey into AI & Data Science here, 90% buzzword-free. Follow me and let's grow together!

Nobody wants to own "data governance".

Why would they?
It's a parade of boring checklists, meetings, and policy docs that nobody reads.

Here's the truth:
Most data governance programs fail because nobody actually uses them. They look tidy in Confluence, but out in the wild, chaos rules. Broken fields linger for months. Data is "owned" by committees where deliverables go to die. Everyone has an opinion, nobody takes action.

If you want traction, skip the dog-and-pony show. Screw the frameworks. Fix the basics.

Start with 3 hard questions:
→ Who touches this data daily, and do they know they're supposed to care?
→ Can anyone remember the last time a policy was enforced, or did it get waved off 'as a favor'?
→ Are you actually tracking outcomes, or just activity for a quarterly slide deck?

Pull one ugly thread. See what breaks. That's where the real problems hide.
Teams don't need another dashboard for governance. They need someone who will show up, call the mess, and fix it when it's right in front of them.

Drop the ceremony, ship actual ownership with clear consequences.
That's governance that works. Otherwise, enjoy your next "policy refresh" meeting.

🔔 Follow me Nick Valiotti for fewer frameworks and more outcomes. 
📘 And for the full approach, check out Your Fractional CDO → https://lnkd.in/ddySGAjR

A data strategy that looks perfect on paper

usually dies in real life.

I’ve seen this pattern across companies of every size.
There’s always a polished deck, five-year vision, clean architecture, immaculate governance, dashboards that could win design awards.

And then you look at how decisions are actually made.
→ Sales still lives in spreadsheets
→ Marketing trusts the ads manager more than internal reports
→ Finance keeps a “shadow model” just in case
And that's not because people hate data, but simply because the strategy never met reality.

From a Fractional CDO standpoint, this is the core mistake:
teams confuse designing a system with changing behavior.

The thing is, data strategy is an operating model. So if adoption isn’t happening, the problem is rarely tooling. It’s usually one of these:
1. No clear decision owner
2. No agreed definition of “good enough” data
3. No incentive to switch from existing workflows
4. No tolerance for messiness during transition

A strategy that only works when everything is clean, aligned, and fully implemented is just a concept. Real data leadership starts smaller and rougher:

– Which decisions actually matter this quarter
– Who owns them
– What data is “good enough” to act
– Where the workflow already lives
If the answer doesn’t fit into how people already work, it won’t be used, no matter how elegant it looks.

Strategy isn’t supposed to be beautiful.
It’s supposed to survive contact with reality.

-------

👋 Hi, I’m Nick Valiotti — Fractional CDO. I help teams turn “data vision” into decisions people actually use. If your dashboards look great but nobody trusts them, that’s usually where I start.

This is what an executive dashboard is supposed to feel like. 

Here’s why this works:

First, the top row does exactly what an executive needs in 15 seconds.
Total Spend. Spend per Vendor. Active Vendors. New Vendors.
You immediately see magnitude and direction. $6.0M total. Down 4.6% vs last year. Vendor count slightly down. No new vendors.

That alone tells a story: consolidation, mild cost compression, no aggressive expansion. You don’t need to hunt for it.

Second, the month-over-month bar chart is clean and comparative.
Current period vs reference period, side by side.
No rainbow colors. No 3D nonsense.

You can instantly spot May’s spike. That’s where a CFO’s brain goes:
“What happened in May?”
That’s the right kind of friction, it pushes conversation.

Third, spend by category is ranked.
Wages at the top. People expenses next. Marketing. Raw material.
This is prioritization.

The business conversation naturally becomes:
Are wages aligned with revenue?
Is marketing spend justified by growth?
Are raw material costs creeping?
Ranking forces trade-offs. That’s strategic.

Fourth, spend by vendor exposes concentration risk.
One vendor at $2.7M.
The next two under $1M.
That’s real leverage.
If I’m running this company, I now know where renegotiation power matters most.

And then there’s the vendor location map.
It’s not there for decoration. It hints at geographic concentration. Supply chain exposure. Regional dependency.
If a disruption hits one of those clusters, what happens to us?

Now, what makes this dashboard strong isn’t just what’s on it.
It’s what’s not.

No vanity KPIs. No metric inflation. No 40 tabs of “just in case.”
This is a spending control cockpit.

If I’m a CEO in this company, I can walk into a board meeting with this and have an intelligent discussion.
If I’m a procurement lead, I know where to focus tomorrow morning.
If I’m a CFO, I see risk, trend, and leverage in one screen.
That’s the standard.

Always great work from Pradeep Kumar G

Crafting data-driven insights and powerful visual narratives. Data Viz Lead | Tableau Ambassador | IronViz Finalist | IIBAwards | 2nd Most Fav Author on Tableau Public

HR data usually answers "what happened." 

This dashboard answers "what do we do next.

Alice Rooney built this Employment Hero dashboard, Tableau's Viz of the Day, and from a business decision-maker standpoint, it earns that recognition.

Here's why it works:

The 288% YoY headcount growth isn't just a vanity metric sitting at the top. It creates a question the rest of the dashboard actually answers: are we scaling well, or just scaling fast?

A leader can immediately check: 
→ Where salary gaps live (avg $68K vs median $55K. That spread matters) 
→ Which departments are carrying performance vs collecting headcount 
→ Whether satisfaction scores hold up as the org grows 
→ If manager layers are earning their comp premium

That last one is subtle but sharp. Sales has the highest median salary. Also the lowest satisfaction score. Operations is the biggest department. Also the lowest performance score. Those two tensions alone are worth a board conversation.

The role breakdown isn't decorative either. When Associates make up 29% of headcount at $40K median and score 5.2 on performance — you've got a workforce composition story hiding in plain sight.

Nothing here feels like it was built to impress a data team. It's built for operators who need to act.

Great work, Alice!

If you want to build dashboards that drive decisions instead of just displaying data, start before you open Tableau. Grab my free Dashboard Canvas interview template: it helps you design around decisions, not visuals → https://lnkd.in/gHBiWM7m

SQL | Data Visualization | Tableau | Data Analytics and Insights

This is what a dashboard looks like when it's designed around decisions, not charts.

Andreea Scintei built this help desk view and the moment you look at it, you stop reporting and start managing.
Here's what it surfaces immediately.

71.7% of tickets are still open. Average 19.3 days without resolution. 32% of tickets have no severity assigned.
That last number is the one that should make leadership uncomfortable. When a third of your workload hasn't been triaged, response time becomes a guessing game and you're dealing with a prioritization process that isn't working.

Then look at satisfaction. 43% unknown.
Nearly half of all support interactions close without capturing a single feedback signal. You think you're managing customer experience. You're actually managing a process gap you never noticed because your dashboard never showed it.

That's what separates this from a reporting screen.
It doesn't just measure activity. It exposes where the system itself is incomplete.

The standard worth holding every dashboard to isn't "does it show data?" It's "does it force a decision?"
If yours doesn't, it's an expensive screenshot.

-----

I built the Dashboard Canvas to help teams design around that question from the start: decision first, then data → https://lnkd.in/du789A95

You can be the best data storyteller in the company and still never be heard.
Hot take: storytelling isn’t the bottleneck anymore. 
Organizational readiness is.

I’ve seen analysts deliver sharp narratives. Clear problem. Clear stakes. Clear next step. No jargon. No fluff.

Everyone nods.
Nothing changes.

For years we blamed this on “poor communication.” So we trained people to tell better stories. And to be fair, they did. The bar is higher now.

But as Brent Dykes pointed out in his piece on narrative tension, resistance isn’t random. It happens when your data collides with the story already living in someone’s head.

If your insight confirms what they believe, you’re a genius. If it challenges a public bet, a KPI they own, or a strategy they championed… you’re suddenly “too simplistic” or “missing context.”

Same data. Different room. Different outcome.

And here’s the uncomfortable part:
Even perfect storytelling can’t override bad incentives.
If bonuses depend on a metric, people will defend the metric. If leaders are punished for bad news, they’ll resist “gut punch” insights. If no one owns the decision, your story becomes an interesting slide deck.

We’ve upgraded storytelling skills. We haven’t upgraded decision culture.
So yes, learn to frame. Learn to read the room. Understand narrative tension.

But also ask a harder question:
Is this organization built to act on truth — or just to listen to it?

-------

👋 I’m Nick. I write about the messy space between insight and action. If you care about making data actually move decisions, follow along.

Author of Effective Data Storytelling | Founder + Chief Data Storyteller at AnalyticsHero, LLC | Forbes Contributor

Everybody talks about "governance" like it's some magic safety net.
But here's the unpleasant reality: if nobody owns the business impact, your governance is hollow.

You can craft all the policies and perform all the reviews you like. Approve the data tables, sign off the ML models, tick every compliance box.
Still doesn't matter.

Because when it comes down to the actual decision, pulling the trigger on something that changes revenue, costs, or risk, that's where responsibility lives or dies.

Slick governance frameworks lull leadership into a false sense of control. Meanwhile, the hard part stays unsolved: 
Who actually holds the bag when the outcomes are messy? 
Who answers for the fire drill, not just the ticket?

This is where most orgs slip. They confuse process for ownership. Accountability gets spread so thin that nothing sticks. A dozen signoffs. Zero accountability. Everybody involved, nobody responsible.

If you want real governance, stop focusing on artifacts and enforcement. Start asking one brutal question: 

who owns this decision when the stakes are high?

Mature companies bake that into muscle memory. Early-stage ones just play paperwork roulette.
Your models are inputs, not absolution. People run the business, people own the risk. Anything else? Corporate fantasy.
Trust comes from knowing whose name is on the line when it matters, not from a template or a checklist.

Keep your framework. But if it doesn't make ownership explicit, you're not protected. You're exposed.
That's the layer most people never build.

------

I’m Nick, Fractional CDO. Follow for practical data strategy that holds up when revenue, cost, and risk are on the line.

Hot take: your company doesn't need a data warehouse.
It needs someone who knows what questions to ask.

I've worked with Series B companies running the full modern stack. Snowflake. dbt. Looker. Fivetran. The works.
Their founders still couldn't tell me what decision the last dashboard actually changed.

Here's the pattern I keep seeing:

Someone sells leadership on "modernizing the data infrastructure" → The team spends a quarter building pipelines → Dashboards multiply → A Monday morning report gets scheduled that nobody reads by Tuesday.

Meanwhile, the actual business questions - should we kill this product line, which customer segment is worth doubling down on, where is the retention leak - stay unanswered.

Because tooling doesn't answer questions. People do.

Signs your data problem is actually a questions problem:

→ You have dashboards but no standing decisions tied to them
→ Your data team gets pulled into requests, not strategy conversations
→ Leadership asks for reports; nobody asks for recommendations
→ You've onboarded three tools in 18 months and clarity hasn't improved
→ Your analysts are busy but you can't point to what changed because of their work

Most companies at Series A/B don't have a data infrastructure problem. They have a prioritization problem dressed up as a tooling problem.

The infrastructure comes second. Always.

The highest-leverage move at that stage is someone who can sit in a leadership meeting, understand what decisions are actually on the table, and build backwards from there. That's what changes the ROI on everything that follows.

More data flowing through better pipes doesn't fix a room that doesn't know what it's trying to decide.

---

💬 Founders: what's the last decision your data team actually influenced? Not reported on. Influenced.

Most dashboards fail for the same reason meetings do: people think they’re talking about the same thing.

Building dashboards is really about building a shared language. Because the dashboard is the sentence everyone reads.
The data stack is the grammar, dictionary, and context that make that sentence mean the same thing to everyone.

Most teams start with charts, filters, or BI tools. That’s backwards.
Here’s the analytics stack, translated into language — in the order it should be built:

0. Question level — intent
What decision are we trying to make? Approve budget, change pricing, pause a campaign?
If you can’t answer that, every chart that follows is cosmetic. Language exists to trigger action, not to sound correct.

1. Source systems — letters
The smallest units of truth.
Users clicked. Money moved. Tickets were opened.
No interpretation yet, just symbols.

2. Ingestion layer — words
How letters become something recognizable: tracking, events, APIs.
If this layer is sloppy, you just get fluent gibberish.

3. Storage layer — the written record
Where words are preserved over time. Warehouses that keep history consistent.
If yesterday’s facts change today, trust disappears.

4. Transformation layer — grammar
The rules for combining words. Joins, aggregations, business logic.
Bad grammar doesn’t just look wrong, it changes meaning.

5. Semantic layer — the dictionary
What the words officially mean: “Revenue”, “Active user”, “Conversion”, “Churn.”
If definitions differ, agreement is an illusion.

6. Analytics layer — the sentence
Dashboards, drilldowns, ad-hoc analysis. This is what people actually read.
A good sentence reduces thinking at the moment of decision. A bad one creates false confidence.

7. Decision layer — action
Where language turns into movement. Operating reviews, alerts, scorecards. If no behavior changes, the sentence didn’t matter.

When teams skip a layer, they don’t get alignment, they get something closer to the Tower of Babel: everyone talking, everyone nodding, nobody truly understanding each other.

Good companies build dashboards.
Great companies build shared language, and make decisions in it.

-------

If this sounds familiar, I turned the framework into a simple dashboard discovery cheatsheet. It’s a ready-to-use Figma canvas you can grab for free → https://lnkd.in/dY4TnMXi

A common question I hear:
“What does a CDO actually do? And why can’t our senior data analyst just do that?”

It’s a fair question.

The thing is:
A data analyst’s job is to answer questions.
A CDO’s job is to decide which questions matter and which ones shouldn’t be asked at all.

Analysts work inside the system.
CDOs are responsible for the system itself.

That means:
→ deciding what data the company should trust
→ setting priorities when everything feels important
→ choosing what to build, buy, or ignore
→ clarifying ownership so decisions don’t stall in debate

None of this is about SQL, dashboards, or tools. It’s about judgment.
At Series A–C, this gap becomes visible. Data exists. Dashboards exist. Talent exists. But nobody is accountable for direction.
That’s usually when founders start feeling the drag: slow decisions, duplicated work, quiet confusion.

This is where a fractional CDO helps — not by replacing analysts, but by giving them direction.

I wrote a short guide explaining how this actually works in practice: what a CDO does, when the fractional model makes sense, and what typically changes in the first 90 days.

Feel free to grab it in my Featured or via the link → https://lnkd.in/dxkdm3nq

👋 I’m Nick, Fractional CDO. If you’re looking for this kind of data leadership, DM me.

Attrition isn’t an HR metric. It’s an operating risk.
This dashboard by Anastasiya Kuznetsova treats it that way.

16% attrition. 237 people gone.
No fluff. No engagement score theater. Just the number that matters.

→ Then it breaks the problem down into levers.
Work-life balance. Income. Tenure.
You immediately see where churn concentrates. Lower balance scores. Lower income bands. Early years at the company.

→ Then the role breakdown.
Sales Representatives at 40%. Lab Technicians at 24%. HR at 23%.
That’s revenue risk. Capability risk. Continuity risk.

And the employee table matters more than people admit.
Because when leadership says, “Show me who,” you can.
No hiding behind averages.

What makes this dashboard strong is alignment.
Every visual ties back to one outcome: retention stability.
So it doesn’t try to be clever. It forces the right conversation.

Before you build another dashboard, map the decision. Use my free Dashboard Canvas to get it right → https://lnkd.in/dnRsTHGx

Hot take: dashboards are how data teams avoid executives, not communicate with them.

Building a beautiful dashboard feels like work. It has colors. It has filters. It has a date range selector nobody asked for. You spent three weeks on it. An exec glanced at it for 11 seconds and went back to their emails.

And this is a comfort zone problem.

Data teams are brilliant at building things that let them never have to be in a room defending a number. Dashboards are the perfect alibi.

"It's all in the report" is the data equivalent of "I sent an email"

The leaders who actually move decisions don't hide behind reports, they show up. Here's a quick gut check on where you actually stand:

How exec-ready is your communication strategy?

→ Do you have a recurring slot with at least one C-suite stakeholder — not to report, to think together?
→ Can you explain any metric on your dashboard in terms of a business decision, without looking at the dashboard?
→ Did you shape the question before the last board meeting — or just answer it?
→ When something shifts in the data, do you call someone or update the report?

If you're mostly updating reports, you're dashboard-first.
Dashboards have their place. But if that's your primary channel with the C-suite, you're not being strategic. You're hiding.

Hat tip to Tris J Burns 🔥 for sparking this one.

-------

♻️ Repost to help someone confusing dashboards with strategy. 
🔔 Follow Nick Valiotti for no-nonsense takes on data leadership and the CDO agenda.

Transform your AI & data ambition into action | xQuantumBlack, xMcKinsey | Global top 100 Innovators in Data & Analytics ’24 | Founder & Director, Cambiq

AI and data are creating some of the most exciting opportunities organisations have seen in years.

With the right approach, they can unlock better decisions, smarter operations and more rewarding ways of working.

I help organisations establish, nurture and scale their use of AI, AI agents and data, turning potential into real outcomes and sustained enthusiasm.

If you’re wondering how you, your business or your team can get up to speed, get started and confidently drive value from AI and data, I can help. 
Whether you’re a large or small organisation, I bring pragmatism, creativity and a strong bias to action to support you to succeed.

If you're wondering how to get started with AI and data, I offer executive education sessions to help bring to life the opportnity for you and your team. We can tailor this to your industry and business.

If you want to do more, I can help you:
1. Develop a clear AI and data strategy that drives growth, efficiency and trust, from early ideation through to business cases and board-ready narratives
2. Create and embed responsible AI and data governance that reduces regulatory risk while making it easier for teams to use data well
3. Build practical adoption strategies so leaders and teams feel confident, capable and motivated to use AI and data
4. Set up a new AI or data program, or turn around an existing one, to maximise the chance of lasting success

I’m the founder and Director of Cambiq, where we work with organisations to turn their own data into commercial and operational advantage through AI strategy, governance, education and delivery.

My background combines McKinsey consulting, senior industry leadership and strong technical foundations. That mix helps me shape strategy, navigate complex stakeholder environments and, most importantly, drive real adoption on the ground.

Inspiring growth and fostering a culture of learning are passions I bring to everything I do. I’m genuinely optimistic about what AI can enable when people are supported to understand it and use it well.

I work flexibly, either on a project basis or as an interim executive embedded within your team. If you’re looking for a partner, advisor or speaker to help your organisation build confidence and capability in AI and data, I’d love to connect.

You can reach me at hello@cambiq.com

Clare Kitching

Everyone talks about AI models.
Very few talk about AI systems.

When you look under the hood of most AI today, you rarely find just a large language model.

You find layers.
Context. Memory. Retrieval. Tools. Autonomy.

This diagram shows the progression.

▶️ Large Language Models (LLMs) are the foundation.
They are massive neural networks trained to understand and generate human language.
They take an input and produce a response based on patterns learned during training.
Powerful, but limited to what they already know.

▶️ RAG (Retrieval Augmented Generation) connects an LLM to external information sources like documents or databases.
Before answering, the system retrieves relevant content and uses it to respond.
This reduces errors and keeps answers up to date, but introduces real engineering work around embeddings, vector search, and document design.

▶️ AI Agents go a step further.
They are systems that can take actions on your behalf to achieve a task.
They plan steps, call tools, execute actions, and track state over time.
This is no longer just about answering questions. It is about doing work.

▶️ Agentic AI systems push autonomy even further.
They coordinate multiple agents, reason over longer horizons, manage memory, and pursue goals with limited human input.
Here the challenge shifts from building capability to orchestrating behaviour.

Each layer demands more than better technology. 
It demands better decisions.
→ Moving from LLMs to RAG means investing in data quality and retrieval design. 
→ Moving from RAG to Agents means trusting systems to act. 
→ Moving to Agentic AI means accepting shared responsibility between humans and machines.

As organisations move up the stack they need to change how they govern, monitor, and own these systems.

Intelligence is the easy part.
Designing for trust is the real work.

Which layer do you think your organisation is actually ready for?

♻️ Repost to help someone understand AI systems.
🔔 Follow Clare Kitching for insights on unlocking value with data & AI.

McKinsey taught me that brilliant people fail when they rush to answer the wrong question.

Once AI enters the room, that temptation gets stronger.
Everyone lines up to press the 'Add AI' button.
Very few stop to define the problem.

“Let’s automate it.”
“Let’s build a model.”
“Let’s try GenAI.”
All solutions.
No clarity.

A vague problem plus AI doesn’t create insight.
It creates noise at scale.

Ask, “How do we make this process faster?”
And speed becomes the goal, even when quality matters more.

Ask instead, “What outcome are we trying to improve for people?”
Now the right solutions have room to emerge.

In my experience, the best leaders are willing to look slow upfront.
They sit with ambiguity.
They argue about the question.
And then they move with confidence.

Before your next AI initiative, try this pause:
Can everyone explain the problem the same way?
Does the problem describe an outcome, not a tool?
Would you still care if AI wasn’t available?

If not, step back define the problem.

AI won’t save you from unclear thinking.
But clear thinking makes AI far more useful.

Where have you seen teams press the wrong button too early?

♻️ Repost to help someone define their problem early.
🔔 Follow Clare Kitching for insights on unlocking value with data & AI.
💎 Get more from me with my free newsletter here: https://lnkd.in/ghBtk6jR

If AI is the answer, what's your question?

46% of employees are using AI tools their company never approved. 
And 32% are not telling anyone.

That’s the reality of AI adoption in most organisations right now.

Personal AI tools are cheap or free and people are experimenting.
Unfortunately, governance hasn’t caught up.

The instinctive response is to write a policy, maybe set up a committee.
But a policy without enforcement is just a suggestion. And most organisations are still at the suggestion stage. 
Some surveys suggest only 25% of organisations have a fully implemented AI governance program.

One way to think about it is that AI guardrails work at four levels. 
And you need to consider all four:

1️⃣  Governance
Set risk appetite and accountability.
55% of organisations now have an AI oversight body. But only 27% have formally added AI to their board committee charters.

2️⃣ Operating model
This is where strategy turns into behaviour.
If a third of your team is hiding their AI use, the policy either doesn’t exist or nobody believes in it.

3️⃣ Process
This is where safeguards live inside workflows with human reviews, approvals and monitoring.
Most organisations underestimate this layer. Yet it is where many real risks appear. 
This layer is the difference between hoping things go well and actually controlling outcomes.

4️⃣ System
Technical safeguards built directly into the tools.
Most enterprise platforms now offer these out of the box. These controls run every time AI is used.

Organisations are usually great at layers 1 & 2, but you can’t stop there.

Start planning how to build all four layers.

Because governance sets the rules.
But systems and processes enforce them.

Think of it like road safety: Signs help.
But guardrails are what stop the crash.

Where do you think most organisations are underinvesting right now?

♻️ Repost to help someone develop their AI governance.
🔔 Follow Clare Kitching for insights on unlocking value with data & AI.
💎 Get more from me with my free newsletter here: https://lnkd.in/ghBtk6jR

One small word is causing big confusion in AI programs:
Agent.

Right now, many AI workflows are being labelled “agentic”.
Even when they're not really.

Unfortunately, once leaders hear “agent”, they assume:
autonomy, decision making and unpredictable behaviour.

And suddenly governance teams go into full risk mode.

So here’s the difference:

▶️ LLM Chatbots
Input and get your response.
Great for answering questions, summarising documents, and drafting text.
Think customer support assistants, internal help bots, or writing copilots.
They don’t take action in systems.

▶️ RPA (Robotic Process Automation)
These are automation scripts, sometimes enhanced with AI in newer systems.
They trigger tasks across systems.
Great for finance processes, reconciliations, and repetitive back office work.
There’s predefined logic rather than planning to reach a goal.

▶️ RAG (Retrieval Augmented Generation)
These combine LLMs with company knowledge.
The model retrieves information from a knowledge storebefore generating an answer.
Great for internal knowledge assistants, policy questions and research support.
The grounding in trusted data sources can lead to more accurate outputs. But they’re still not agentic because they don’t plan or act.

🤖 Agentic AI
Now we reach the real thing.

Agents can plan steps to reach a goal, choose and use tools, coordinate with other agents and adapt based on feedback.
They behave more like a problem solver than a responder.
Best for complex, multi step work.

Every pattern has value.
But they require different governance, risk controls, and expectations.

If we call everything an agent, we create unnecessary fear and slow down adoption.

Sometimes a simple architecture is exactly what you need.

Where do you see the biggest misunderstanding about agentic AI right now?

♻️ Repost to help someone match their workflow to the right risk.
🔔 Follow Clare Kitching for insights on unlocking value with data & AI.
💎 Get more from me with my free newsletter here: https://lnkd.in/ghBtk6jR

Thank you to Rakesh for the image inspiration.

CEO: “Why aren’t we using AI yet?”
Data team: “We’ve been using it for years.”

Both are right. 
Both are talking about completely different things.

The exec sees ChatGPT writing marketing copy.
The data scientist sees a churn prediction model in production.

The difference matters, because each brings new opportunities, new data needs, and new risks.

AI isn’t one thing.
It’s a spectrum:
From predicting & classifying → to creating → to acting.

Before investing in the next AI pilot:
▶️ Start by understanding whether you need the analyst, creator or executor - maybe it's one, maybe it's a combination
▶️ Then ask: what does it need to be successful and what risks do we need to manage?

Because when everyone understands what type of AI you’re using, expectations align and investment decisions make sense.

Five minutes now could save you months of misaligned pilots and budget debates.

How does your organisation currently perceive AI?

♻️ Repost to help teams agree on what they mean by AI.
🔔 Follow Clare Kitching for insights on unlocking value with data & AI.

3 roles of AI every leader should understand

In 12 months, you’ll wish you started learning AI today.

We’re already in March.

If AI is anywhere on your 2026 goals list, this is your sign to stop waiting.

Google has made it easier than most people realise with its free AI courses.

Google Skills has a whole range of free courses:
▶️ From beginner to advanced
▶️ From multi hour deep dives to 10 minute refreshers
▶️ From videos to hands on labs

Google has been stacking this library for a while now, quietly adding fresh courses along the way.

I’d skimmed it in the past. This time I got properly pulled in by how much sits there, ready to use.

They’re built by experts.
They’re practical.
And they’re available right now.

The full catalogue is here: https://www.skills.google/

To save you time, here are 7 free courses to get you started:

📌 Save this. You’ll thank yourself later.

1️⃣ The Generative AI Leader
The right starting point. Your pathway to understanding capabilities of generative AI and understand how they benefit organisations.
🔗 https://lnkd.in/g-wEa5NY

2️⃣ Agent Fundamentals
Understand the fundamentals of AI Agents, how they differ from LLMs, and where they add value in the real world. A theoretical base before you start building.
🔗 https://lnkd.in/gf5cr-rs

3️⃣ Responsible AI
How to operationalise responsible AI in your organisation.
🔗 https://lnkd.in/ghsJ8FU2

4️⃣ Unlock insights with NotebookLM
Go from overload to insights with NotebookLM helping you get through large volumes of research. Useful if reading reports is part of your job.
🔗 https://lnkd.in/gXQhyv98

5️⃣ AI Boost bites: Supercharge your research (under 10 mins)
Pro tips for conducting thorough competitive research with Gemini Deep Research
🔗 https://lnkd.in/gd-NGSz8

6️⃣ AI Boost Bites: Animated Charts with Gemini (under 10 mins)
Data & animation - a winning combination! Create an animated bar chart race.
🔗 https://lnkd.in/gmvNJkqu

7️⃣ Vibe Coding for Beginners
Get into vibe coding by building a functional web app without code. Worth it if you've had a product idea but no technical background to act on it. Anyone can do this!
🔗 https://lnkd.in/giMXQ_yq

They’re free, self-paced and there are so many more to explore.

Pick the one that maps to your biggest time sink right now and start there.

♻️ Repost to help someone get ahead with AI.
🔔 Follow Clare Kitching for insights on unlocking value with data & AI.
💎 Get more from me with my free newsletter here: https://lnkd.in/ghBtk6jR

Every project starts with hope.
Every timeline ends with a sigh.

Why? Because AI & data projects are the perfect storm of:
→ Optimism bias (a.k.a. the planning fallacy)
→ Fuzzy scope (“we think we want this…”)
→ Unknowns (it’s the first time for everyone)
→ Data, security & ethics curveballs

And this is where the frustration really sets in.

I’ve been there. 
The neat plan, the polished estimate, the stakeholder nods… 
only for reality to unravel in predictable ways.

You imagine a smooth run to delivery.

Reality looks more like:
→ The “clean” data is chaos.
→ Compliance shows up after 3 months.
→ Pilots uncover 20 edge cases you never scoped.

System works, but business team has no idea how to use it.

You start with a project plan.
You finish with a war story.

And somehow… both were inevitable.

But next time you’ll try to
→ Plan with plenty of buffer for the knowns and unknowns
→ Scope what’s in as well as what’s out
→ Adapt again, keeping everyone close as you do

What’s the most unexpected twist you’ve seen in a project?

♻️ Repost to help someone get their project back on track.
🔔 Follow Clare Kitching for insights on unlocking value with data & AI.

How we think an AI project will run vs. How it actually runs

CEO: “Why aren’t we using AI yet?”
Data team: “We’ve been using it for years.”
Both are right.
The problem is they’re picturing completely different things.

AI isn’t one thing, though it usually plays one of three roles:

1️⃣ The analyst 
Traditional AI helps you predict, classify and make better decisions

2️⃣ The creator 
Generative AI helps you draft, summarise and generate ideas quickly

3️⃣ The worker 
Agentic AI helps you execute tasks across systems with less manual effort

That distinction sounds basic.
But I think it's where a lot of organisations get stuck.

They jump straight into tools before they are clear on the job.
And when that happens, expectations get fuzzy very quickly.

The other reason this matters is maturity.
Traditional AI is largely proven. 
Generative AI is moving fast. 
Agentic AI is emerging powerfully, and still needs tight guardrails.

So the leadership challenge is both understanding what each type can do and understanding what each type needs:
Clean data. Clear rules. Strong controls.

That’s how you move from experimentation to value.

Which of the three do you think will create the most value over the next 12 months?

♻️ Repost to help someone get their idea into action.
🔔 Follow Clare Kitching for insights on unlocking value with data & AI.

3 Roles of AI Every Leader Should Understand

2026 is the year AI stopped being a feature.
It became part of the operating system of the enterprise.

Across the latest 2026 flagship reports, one message is clear: this is the inflection point for enterprise AI.

Despite different perspectives, they converge on 3 strategic signals:
1️⃣ Agentic AI is shifting from experimentation to enterprise infrastructure.

2️⃣ Value comes from redesigning work, not layering automation onto legacy processes.

3️⃣ Strong data foundations and governance enable speed rather than slow it down.

This is less about adopting technology and more about redesigning how the enterprise actually runs.

For any executive planning the next 12–24 months, these are the anchor reports:
1/ BCG: AI Radar 2026
🔗 https://lnkd.in/gckyZdsV
AI transformation is shifting from a CIO-led initiative to a CEO-led mandate.

2/ IBM Institute for Business Value: The Enterprise in 2030
🔗 https://lnkd.in/g_u9d8t4
Five bold predictions for the "smarter enterprise" of 2030.

3/ McKinsey: Global Tech Agenda 2026
🔗 https://lnkd.in/gJGsqpkT
Why top CIOs are evolving from cost managers to strategy architects who rewire enterprises for growth.

4/ Deloitte: State of AI in the Enterprise 2026
🔗 https://lnkd.in/gE3-e9XT
Exploring the edge of AI adoption, including the rise of sovereign, agentic, and physical AI.

5/ World Economic Forum: Proof over Promise
🔗 https://lnkd.in/gQtVQN8t
A practical anchor on how organisations are scaling AI and turning it into outcomes.

6/ Microsoft WorkLab: Agents are here: Is your company prepared?
🔗 https://lnkd.in/g5TXhQyW
A readiness & adoption focused on people, process, culture, and governance for agentic AI rollout.

7/ Google Cloud: AI Agent Trends 2026
🔗 https://lnkd.in/gU7ucWye
A trends report mapping how AI agents are being used across industries.

8/ Accenture: The New Rules of Platform Strategy in the Age of Agentic AI
🔗 https://lnkd.in/gpreVptN
A playbook for reinventing platform strategy to align people, systems, and agentic AI for growth.

9/ McKinsey: The State of Organizations 2026
🔗 https://lnkd.in/gVb5mbkr
A foundational lens on whether your organisation is structurally ready to absorb AI, automation, and continuous change.

Now it’s time to consider how deeply you’re willing to redesign the enterprise around AI.

These decisions compound.
The workflows you hand to agents, the data foundations you build, and the controls you put in place will shape your performance for years.

♻️ Repost to help someone get their AI strategy together.
🔔 Follow Clare Kitching for insights on unlocking value with data & AI.

If AI feels both everywhere and strangely unclear, you’re not behind.
You’re normal.

Most execs I speak to aren’t confused about AI.
They’re confused by the language.

The same terms get thrown around in boardrooms, vendor decks, and strategy papers. Often interchangeably. Often vaguely.

So here’s a plain English guide to the AI terms you’re hearing most right now.
No tech background required.

▶️ Generative AI (GenAI)
AI that creates new things. Text, images, music, code.
This is usually what people mean when they say “AI” today.

▶️ Large Language Model (LLM)
A very large model trained on vast amounts of text to understand and generate language.
This is the engine behind tools like ChatGPT.

▶️ Foundation Model
A large, pre trained model that can be reused for many tasks.
Most of the real power and cost in AI sits here.

▶️ Agentic AI
AI that can plan, decide, and act to achieve goals.
More autonomy means more value, but also more risk if not governed.

▶️ AI Agent
An AI that behaves more like a worker than a chatbot.
It can reason, use tools, and take actions toward a specific objective.

▶️ Multi Agent System
A group of specialised AI agents working together.
Many agents together often scales better than one general purpose agent.

▶️ RAG (Retrieval Augmented Generation)
A way to improve AI accuracy by pulling in up to date or trusted external information before responding.

▶️ Prompt Engineering
Designing instructions to guide how an AI behaves.
Useful today, but becoming less critical as models improve.

▶️ Context Window
How much information an AI can consider at one time.
Many failures come from hitting this limit.

▶️ Tool Use
The ability for AI to interact with other systems like software and APIs.
This is what turns AI from interesting to genuinely useful.

▶️ Human in the Loop
A safety approach where humans review or approve high impact AI actions.
One way to keep control without killing automation.

▶️ Hallucination
When AI confidently makes things up.
Confidence does not equal correctness.

If you’re making decisions about AI investment, risk, or strategy, shared language matters more than hype.

What AI terms do you see most confusion coming from?
What other terms would you add as essential?

♻️ Repost to help someone get their idea into action.
🔔 Follow Clare Kitching for insights on unlocking value with data & AI.

Your biggest data risk isn’t AI.
It’s the layers no one wants to fund.

“We already have the data.”

I hear this a lot. And it’s usually only half true.

Yes, the data exists. 
But turning it into something useful takes more than plugging in a tool.

This image, inspired by the wonderful Marcel Dybalski, shows a modern data stack. 

▶️ At the bottom are the systems that create data. 
Built to run the business, not explain it.
This is your CRMs, ERPs, apps and sensors.

▶️ Next comes ingestion. 
Getting data out safely and regularly, without breaking anything.

▶️ Then storage. 
Where data lives, how it’s structured, and how long it’s kept.

▶️ Above that is processing and orchestration. 
Running jobs in the right order. Handling failures. 
Making sure the data actually arrives when it should.

▶️ Then curation and transformation. 
Cleaning data, shaping it, and agreeing on what the numbers really mean.

▶️ Next serving.
Making data fast and easy to reuse.

▶️ Only after that do you get the visible layer. 
Dashboards. Self service reporting. AI. Insights.

▶️ And across all of it sit security and governance. 
Invisible when done well. Painful when ignored.

If you’ve ever wondered why data work takes time, this is why.

Data isn’t magic. 
It’s layers of careful work stacked on top of each other.

The least visible work is often the most important.

If you want data you can trust, you have to respect every layer, not just the top one.

Which part of this do you wish more people in your organisation understood?

♻️ Repost to help someone explain their data stack.
🔔 Follow Clare Kitching for insights on unlocking value with data & AI.
🐨 Get more from me with my free newsletter here: https://lnkd.in/ghBtk6jR

Two ways to get AI wrong:
chase every shiny object, or play it so safe you sleepwalk into irrelevance.

Some organisations pile into agents and GenAI and hope it changes everything.

Others stay in their comfort zone. 
Rules. Reporting. Incremental improvements.

Both create hidden risk.

If you only chase big bets, results get fragile fast. 
If you only back safe bets, you slowly fall behind.

The AI strategies that hold up over time look more like portfolios.

You have dependable layers:
▶️ Clear rules that protect compliance. 
▶️ Predictive models that quietly improve margin and retention. 
Systems you can actually explain when something goes wrong.

And you have a few deliberate growth bets.
▶️ Deep learning where the problem is genuinely complex. 
▶️ Generative AI where speed and leverage matter. 
▶️ Agents where autonomy creates advantage and you have the controls in place.

Different layers. Different risk. Different oversight.

The foundations should fund the experiments. 
The experiments should earn the right to scale.

AI is a set of choices about where to take risk and where to stay predictable.

If your board asked you to explain your AI mix like an investment strategy, could you justify the balance?

♻️ Repost to help someone find their righ AI portfolio.
🔔 Follow Clare Kitching for insights on unlocking value with data & AI.
💎 Get more from me with my free newsletter here: https://lnkd.in/ghBtk6jR

AI isn’t the strategy.
Business fit is.

This is something I wish more leaders understood:
AI gives you options. And each one creates value in a different way.

Yet many organisations jump straight to the latest. Agents. GenAI. Big bets.
They skip the simpler choices, where value often hides.

▶️ Rules and logic create certainty. Useful in a highly regulated environment.

▶️ Predictive models give foresight. They’ve been used to drive up revenue for years.

▶️ Deep learning gives perception. Computers can spot defects humans miss.

▶️ Generative AI creates. This is the public moment for AI.

▶️ Agents execute. Software that behaves less like a model and more like a junior employee.

The opportunity lies in knowing which option fits your problem.

AI isn’t one decision. It’s a portfolio you actively manage.

Start with the simplest approach that works for your problem. Add complexity only when it earns its place.

If more AI programs followed this logic, we’d see fewer pilots and more impact.

Which options do you think are usually overlooked but add great value?

♻️ Repost to help someone understand their AI options.
🔔 Follow Clare Kitching for insights on unlocking value with data & AI.
☘️ Get more from me with my free newsletter here: https://lnkd.in/ghBtk6jR

AI tools are starting to feel like a holiday buffet.
But how much can you take?

Everything looks great. 
Everything feels worth trying.

So you take a little bit of everything.

But you never really enjoy any single dish.

The same thing happens with AI.

If you constantly jump between tools, you stay at the surface and never build real skill.

The biggest productivity gains I’ve seen come from people who do the opposite:
They choose a small set of tools. Sometimes just one.

Then they go deep.
They test prompts, refine workflows and automate small pieces of work.
Over time it compounds.

Sure, if something new and special appears, they give it a try.
But they keep a solid core.

Getting results from AI is more about building habits than collecting tools.

If you had to choose just one AI tool to keep, which would it be?

♻️ Repost to help someone focus their AI habits.
🔔 Follow Clare Kitching for insights on unlocking value with data & AI.
💎 Get more from me with my free newsletter here: https://lnkd.in/ghBtk6jR

AI tools are endless. Value comes from using them.

If your AI strategy feels underwhelming, you might be missing a piece.

Not a bigger model.
Not more data.
A different slice of value.

Right now, many organisations are betting heavily on one aspect of AI.
Usually Generative AI.
It’s visible, exciting and demos well.

But AI helps create business value in many ways.

→ Rules and logic handle clear yes or no decisions. 
→ Predictive AI tells you what is likely to happen. 
→ Deep learning handles complex pattern recognition. 
→ Generative AI communicates and creates. 
→ Agents execute across workflows.

Each plays a distinct role.

When one is overlooked, performance suffers and value is left on the table.

Too often I see teams zoom in on a narrow slice of AI and miss the broader set of proven use cases available to them.

When you step back and design the right combination, the impact compounds:
Costs decrease. 
Decisions improve. 
Customer experience lifts.

AI works best as a portfolio, not a single bet.

If you mapped your current investments across these pieces, where would the gaps show up?

♻️ Repost to help someone get their value from AI.
🔔 Follow Clare Kitching for insights on unlocking value with data & AI.
🍨 Get more from me with my free newsletter here: https://lnkd.in/ghBtk6jR

Many companies say they want Agentic AI. 
But most conversations skip the fundamentals.

Companies want autonomous agents, AI employees and work done end to end. Sometimes, it sounds like magic...or mayhem.

But when you strip away the buzzwords, agentic AI isn’t magic.
It’s a set of design patterns and capabilities that allow AI systems to plan, reason, act, and improve over time.

If you want to understand how AI agents work, these 15 concepts are the building blocks.

Here’s a quick guide you can bookmark:

▶️ Agent
An AI system that can pursue goals, take actions, and adapt to results.

▶️ Tool use
An agent’s ability to use external tools like APIs, search, databases, or code.

▶️ Planning
Breaking a goal into a sequence of steps to enable multi step work.

▶️ Reasoning
Analysing information, comparing options, and choosing the best next step.

▶️ Memory
Information an agent retains within sessions or across tasks.

▶️ Guardrails
Rules and checks that constrain agent behaviour and keep it safe.

▶️ Human in the loop
Human review or approval during key steps.

▶️ RAG (Retrieval Augmented Generation)
A technique that pulls relevant external knowledge at runtime.

▶️ Context window
The amount of information a model can handle at once.

▶️ State
The current status of a task, workflow, or environment.

▶️ Evaluation
Measuring an agent’s performance, accuracy, reliability, and behaviour.

▶️ Orchestration
Managing the coordination between agents, tools, and workflow steps.

▶️ Autonomy spectrum
The range from fully human directed systems to fully autonomous agents.

▶️ Execution loop
The cycle of observe, decide, act, and evaluate.

▶️ Observability
Monitoring what agents are doing, why they are doing it, and how well they perform.

Agentic AI is the combination of these capabilities working together.

If you work in data, AI, or digital transformation, understanding these concepts will soon be table stakes.

Save this for later.

Which of these concepts do you think most teams underestimate today?

♻️ Repost to help someone get their idea into action.
🔔 Follow Clare Kitching for insights on unlocking value with data & AI.
🍨 Get more from me with my free newsletter here: https://lnkd.in/ghBtk6jR

If your AI stack vanished tomorrow, what happens?


I used to dismiss Sovereign AI as tech nationalism.
Talk of building local models felt political, rather than practical.

But having worked in critical infrastructure and heavily regulated sectors, I have a new appreciation.

When your systems underpin payments, energy or essential services, resilience isn’t a buzzword. It’s a necessity.

Sovereign AI is about building options to strengthen resilience and benefit from the economic upside of AI (and whatever comes next).

Options to:
🔹 Keep sensitive data under your jurisdiction
🔹 Develop local skills and IP
🔹 Blend global innovation with local control
🔹 Stay compliant and build trust

The numbers are hard to ignore:
→ 70 percent of leading models still come from the US [Accenture]. 
→ 71 percent of leaders now call sovereign AI a strategic concern [McKinsey]. → The market is projected to head towards $600 billion by 2030 [McKinsey].

This calls for pragmatism from organisations.
1️⃣ Choose what needs sovereignty. Accenture says a good rule of thumb is about a third of AI workloads need sovereignty.

2️⃣ Mix local and global.

3️⃣ Make governance operational, not theoretical.

I used to think this was politics.
Now I view it as risk management and long term value capture.

Is sovereign AI strategic necessity in your sector, what opportunities do you see?

♻️ Repost to help someone get their idea into action.
🔔 Follow Clare Kitching for insights on unlocking value with data & AI.
💎 Get more from me with my free newsletter here: https://lnkd.in/ghBtk6jR

The hardest part of learning AI isn’t the tool.
It’s giving yourself permission to experiment.

This International Women’s Day, the theme is Give to Gain.
Here’s one thing worth giving yourself:
Permission and time to build your AI skills.

Too many capable people are still waiting for the perfect moment.
A course, certification or someone telling them they’re ready.

Instead, start small. Take 10 minutes to try a tool on something you already do. But don't stop there...

Then give yourself a few hours over the next week or two to move through the real learning curve:
→ The high of the first success.
→ The frustration when the output is almost right but not quite.
→ And eventually, the moment where you start to find your rhythm.

That’s when you start to bust the myths that the wonderful Georgie Hubbard shares in the visal below:
🔺 You don’t need to learn to code.
🔺 You’re not being replaced.
🔺 And it’s definitely not too late to start.
You’re learning how to use AI to strengthen the skills you already have.

The second thing worth giving is what you learn.

One of the things I find most fascinating about AI is how differently people use it. Just talking about it can unlock completely new ways of working.

A big thank you to Nick Zervoudis, who recently opened my eyes to a new way of working with Claude Code (and it has nothing to do with coding).

In the spirit of Give to Gain this International Women’s Day:
1️⃣ Give yourself permission to start.
2️⃣ Give yourself time to get through the messy middle.
3️⃣ Give someone else a hand by sharing what you’ve learned.

Then you’ll gain powerful perspectives:
A clearer view of the opportunities ahead.
A better understanding of AI’s limitations.
And the confidence to position yourself for what comes next.

Because we’re still at the very beginning.

What’s one AI tip or workflow someone shared with you that changed how you work? Share it and give a shout out to the person who has helped you learn.

♻️ Repost to help someone get their idea into action.
🔔 Follow Clare Kitching for insights on unlocking value with data & AI.
💎 Get more from me with my free newsletter here: https://lnkd.in/ghBtk6jR

taplio

The Battle

Nikolay Valiotti

Clare Kitching

Followers

Posts

Engagements

Likes

Performance ScoreNikolay Valiotti's LinkedIn Performance

Performance ScoreClare Kitching's LinkedIn Performance

Ready to grow your LinkedIn presence?

taplio

The Battle

Nikolay Valiotti

Clare Kitching

Followers

Posts

Engagements

Likes

Performance Score.css-goub6s{font-size:var(--chakra-fontSizes-md);font-weight:500;color:var(--chakra-colors-gray-600);display:block;margin-top:var(--chakra-space-1);}Nikolay Valiotti's LinkedIn Performance

Performance ScoreClare Kitching's LinkedIn Performance

Ready to grow your LinkedIn presence?

Performance ScoreNikolay Valiotti's LinkedIn Performance