Diagnostics

Count the cards before you bet.

Every strategy is a bet on the organisation that has to deliver it. A TalentOptima diagnostic shows you the hand you are actually holding: one ambiguous organisational question turned into evidence, a decision your leadership team has committed to, and a baseline you can re-measure a year on.

I built the engine myself, AI-native from the first line, and I deliver every diagnostic personally.

My experience in building AI-powered insights-rich diagnostics

I have deep experience in employee and organisational feedback tools.
Manager90, the manager-feedback tool deployed at Nike and Sanofi, and the simplification diagnostic behind the Volcano programme. In both cases I designed the questions, chose the audiences, and lived with what the answers did to the organisation afterwards.
My team was accountable for group people analytics at Sanofi.
I’ve had full accountability for the solution, from its design and the end-to-end journey to technology partner selection, deployment and adoption.
I have deep product and engineering expertise building production-grade applications with AI.
This engine is one of them: the respondent’s journey, the diagnostic’s own journey through the seven steps, and analysis that finds qualitative and quantitative patterns across a body of evidence that used to need a team and a quarter. This site is another, and you can inspect how it was built.
My education combines a science masters and an MBA from IMD.
An MSc in physics and an MBA from IMD. Statistics, sampling and modelling on one side; how a business actually runs on the other. An instrument built without the first is unreliable, and one built without the second measures things nobody will act on.

The method

Planned backwards from the decision it has to serve.

Collecting the data is one step of seven, and the easiest. The work that makes a diagnostic worth acting on happens either side of it. Before anyone is asked anything, I work backwards from the decision: what do we need to know, who can tell us, and what will the questions themselves signal? Afterwards, AI helps me read across the evidence, find patterns and contradictions, and work through the options. Your leadership team can then challenge the findings and decide what to do.

Before you askThe decision, the instrument, and the signal the instrument sends.Steps 1, 2, 3
When you askOne step, and the smallest part of the work.Step 4
After you askEvidence into synthesis, synthesis into a decision, the decision into a loop.Steps 5, 6, 7

Before you ask

Frame the end-state decisions.
A diagnostic exists to serve a decision, and a decision starts life as a hypothesis. So the first move is to name the decision, write it as a hypothesis, and work out the direct and indirect questions that are genuine inputs to it. Centralise or decentralise? Behind that sit five to seven questions that actually bear on the answer. Inputs are mapped to hypotheses, not collected and then searched for meaning.
Validate the inputs and the audiences.
Every question is then matched to the population that can answer it. Not every audience holds an equally validated view, and some know some things far better than others, so weighting is designed in rather than every response counting the same. This is the step that separates a diagnostic from a staff survey.
Manage the messaging.
The questions are already part of the change. Ask someone to rate a colleague against twenty leadership behaviours and you are showing them what leadership means here. They may have started with a different view. Now they know what matters, and they are likely to notice those behaviours more closely next time. Alongside that is the Hawthorne effect: knowing you are being observed can itself change how you behave. So a survey does more than collect answers. It can build understanding and prepare people for change, or create anxiety if it arrives without explanation. I design the questions and the communication together, so we use that opportunity deliberately.

When you ask

Collect the data.
Respondents are mapped against employee data or another source, and two things are settled before a single question is fielded: whether responses are anonymous, and what is held against a respondent either way. Business unit, tenure, level, job family, geography. The cuts you will want in the next step are decided here, not discovered later.

After you ask

Assess and synthesise.
A qualitative and a quantitative pass, cut in the ways that inform the decision rather than the ways the data happens to fall: where value is created, delayed or lost. Every finding stays traceable to its source, its date and its strength. Scores carry their confidence separately, so you get a rating you can interrogate rather than a single number that hides the variation, and disagreement is recorded rather than averaged away.
Align and commit.
The synthesis becomes input to the decision named in step one, and how it arrives is your choice: the evidence alone, presented up front with the discussion after it, or the evidence plus scenarios, with the options set out and the case for and against each. Then one workshop, where your leadership team challenges the findings, makes the trade-offs and commits.
Pulse check on progress.
With a baseline taken and a direction chosen, name the shift you expect against it: what should go up, what should go down. Then design the re-measure. A pulse can be a subset of the questions, or the full assessment to a subset of the population, chosen by where impact or resistance is most likely. These are feedback loops that update the work as it runs and check that the value behind the decision is actually landing.

Areas of focus

Available diagnostics to inform the work

Four areas of focus, each built around a question you may be facing. They use the same engine and can combine: evidence is collected once, and where your question spans two areas, I look at it through both.

AI value-to-work

Where can AI redesign work and increase value, not merely automate activity, and what has to change in workflows, roles, capability and governance to realise it?

Learn more

Who asks: CEOs and COOs whose AI investment is not yet changing the work; CHROs asked what the workforce has to become; PE sponsors modernising a portfolio.
You receive: A value-to-work map: where AI augments the work, changes it, or should not touch it.
The priority workflows and roles, with the evidence and confidence behind each.
Capacity and capability stated as hypotheses, never as headcount targets.
A ninety-day backlog with owners, gates and value measures.

Transformation health

What is stopping the transformation moving at the pace the strategy needs, and what has to be reset in the system before more activity is added?

Learn more

Who asks: CEOs, CFOs and boards who want an independent view; transformation sponsors; PE operating partners.
You receive: A transformation-health profile with the strength of the evidence and the key contradictions.
Root causes separated from symptoms.
Stop, start and continue decisions, and a prioritised reset backlog.
A blueprint for the transformation engine: governance, roles, cadence and escalation.

HR function reinvention

What value must HR create for the business now, and how should its work, capability, operating model and use of AI change to deliver it?

A worked example of the thinking: The first 90 days as CHRO, a practical guide (PDF, seven pages, opens in a new window)

Learn more

Who asks: Group CHROs and CPOs, often with the CEO or CFO; incoming people leaders who need an evidence-backed view of what they have inherited without auditing the team.
You receive: A from-to view of the HR value proposition and operating model.
The priority value streams and the friction in them.
An AI-enabled redesign hypothesis for the work and the capability behind it.
A co-created roadmap with sequence, ownership, measures and early wins.

For a new CHRO or CPO it runs discreetly: documents, external evidence and the leader’s own listening first.

Organisational friction and simplification

Where is complexity consuming capacity, slowing decisions and weakening value for customers and employees, and what should be removed, redesigned or reallocated?

Learn more

Who asks: CEOs, COOs and CHROs under cost or speed pressure; leaders whose growth has created fragmentation; organisations where AI has freed capacity the structure has not used.
You receive: A map of where friction destroys value, following the work across boundaries rather than starting from the org chart.
Layers, spans, hand-offs and decision rights against observable criteria.
The interventions prioritised: structural, process, behavioural and AI-enabled.
A simplification path led by value, not an across-the-board cut.

What you get

Evidence you can interrogate, priorities you can act on

Whichever question we start with, you leave with:

A clear statement of the decision, what it covers and what it does not.
An evidence map and scores for each dimension, with the gaps and confidence behind them.
A view of what is working, what needs to change and the value at stake.
Priorities ordered by value, effort and speed, with a roadmap and named owners.
A baseline and a plan for measuring again, so you can see whether the changes are working.

Every rating shows the evidence that supports it and the evidence that is missing. Disagreement is recorded, not averaged away. You can challenge any fact; the diagnostic judgement stays independent.

Your team can repeat the assessment without me in the room. Where clients want it, that can also mean a licensed re-survey across a portfolio or a group. Sustainable without me, by design.

Built, not bought

The intersection, in practice

This is the clearest example I have of where organisational judgement and hands-on AI building meet, because I built the engine myself. AI does the work it is good at: ingesting the evidence, transcribing and retrieving it, synthesising across sources, detecting contradictions, keeping every finding traceable, and drafting scores for review. I do the work that has to stay human: framing the decision, applying twenty years of pattern recognition across four industries, judging the exceptions, and owning every recommendation that reaches your leadership team.

Because the evidence, analysis and baseline are held together in the engine, I can return to the same question a year later and show what has changed without reconstructing the work.

Why it matters now

How organisations find things out is being rebuilt

I have spent much of my career working with annual surveys, rounds of interviews and findings that took weeks to pull together. Too often, by the time the findings arrived, the question had moved on.

That is changing underneath us. Evidence that was unreadable at scale (documents, transcripts, tickets, the free text nobody had time to code) can now be read in full and across sources, so contradictions surface instead of averaging out. Interviews can be run with modern AI practice, so depth stops costing a diary slot and a hundred people can be asked properly rather than twenty. Listening moves from an event to something closer to continuous.

The risk is just as obvious. Faster and cheaper evidence is not better evidence, and a model that reads everything will find patterns that are not there and describe them confidently. Which is why the framing, the judgement and the decision stay human, and why the method matters more now than it did when the instrument was slow.

Questions

The questions I am always asked

How disruptive is this for my leadership team?
Less than most people fear, and by design. The evidence is mostly collected without pulling your leaders into a room: existing strategy, operating-model and performance documents, the data you already hold, short asynchronous questionnaires, guided voice or written responses, and a small number of live interviews where the subject is sensitive or senior.
Do we have to run workshops?
One, at the end, and it is the point of the whole exercise: your leadership team challenges the findings, makes the trade-offs and commits to action. Group workshops are deliberately kept out of evidence collection, because they produce negotiated narratives before the evidence has been understood.
How long does it take?
A focused diagnostic runs over a few weeks. A broader one, across more than one area of focus or a larger population, takes longer. The collection plan uses the minimum evidence the decision needs.
Is it anonymous, and what do you hold about respondents?
That is settled with you before anything is fielded. Direct personal data is stripped, and results are reported by segment only above an agreed minimum sample size. The attributes held are the ones the analysis will need: business unit, tenure, level, job family, geography.
Do you rank or profile individual employees?
No. The diagnostic assesses the organisational problem. It does not rank, rate or profile named employees.
Does our data get pooled or benchmarked against other clients?
Not unless you expressly opt in, and not until there is a cohort worth comparing against. Your data stays yours and stays isolated; it enters no shared pool. Until then, external material is used as reference points and named as such.
Is the method scientifically validated? Is there a benchmark?
Two things I will not claim. The method is evidence-informed, expert-reviewed and pilot-tested; I do not call it scientifically validated before the evidence exists. And there is no proprietary benchmark until there is a defensible cohort; I will not imply a percentile from a handful of clients.

Bring me the question you cannot yet answer with evidence.

Start the conversation