Nobody Will Ask You To Diagnose Your Talent System. Do It Anyway.

Part 1 of two. Not how to do talent management, but how to read the one you have joined.

Two organisations. Both profitable, both leaders in their sector.

The first promotes by grade and tenure. Nobody’s increment differs from anybody else’s by more than three percent. Development is open to everyone. People stay fifteen years.

The second runs up-or-out. The gap between what its best performer takes home and what its median performer takes home is four-fold. Half the intake is gone in three years, and that is deliberate.

Which one is better at talent management?

Most people answer in a few seconds, and the answer says more about where they have worked than about either organisation. People from the first kind call the second brutal. People from the second call the first complacent. Both are describing the water they swim in.

The question has no answer because it is a poor question. A better question is, what determines the talent management philosophy of an organization?

What this piece is for

This article is about how to read the talent system you have joined. Which philosophy it is running, why it is running them, what it could not change even if it wanted to, and which of your good ideas were dead before you had them.

That skill is not taught anywhere, and nobody will ask you for it. You will be handed a piece of HR – recruitment, or learning, or performance management, or a centre of excellence – and given targets related to just that piece. The assumption is that over a period of time you will work out how the whole thing fits together by absorption: by watching, by making expensive mistakes, by having a proposal die for reasons nobody explains. Most people do get there. It takes most a couple of years and they pay for it in credibility they cannot get back.

You can do it deliberately instead, within your first month in a new organization.

Talent management philosophy is downstream of business strategy

Go back to the two organisations and work backwards. What would have to be true about a business for the first model (tenure based promotions, roughly similar increments) to be correct?

If labour is abundant. If margins are thin. If what you sell is scale and reliability rather than individual brilliance. If no single person moves the numbers enough to notice. Under those conditions, paying everybody roughly the same is not timidity – it is arithmetic. Differentiating pay when nobody is differentiating output is cost with little return and lots of resentment.

Now the second. If revenue sits inside a small number of relationships. If the gap between your best and your median really is four-fold. If the people you would lose can start at a competitor on Monday. Under those conditions, egalitarian reward means you are taking it from the people who generate your margin and giving it to the people who do not.

The philosophy follows the business. It follows the sector, the model, what the organisation competes on, and the labour market it hires from. It does not follow from values, and it does not follow from how modern the HR function is.

Why seven dimensions of Talent Management, and why these seven

There is no agreed definition or scope of talent management. I checked. Researchers of the field have been saying so for twenty years: the term is used to mean succession planning or high-potential programs or simply the whole of HR with a better name – the boundaries move depending on who is speaking.

That is a problem for researchers. For a practitioner it is refreshingly freeing, because it means you can stop hunting for the correct definition and build a list that actually helps you read an organisation.

When you think about talent in a real organisation, you are thinking about four things about your people: who is even counted as talent, where capability comes from when you need it, what people get in return for what they contribute, and how they move to where the work is.

Then two things about your systems – not about people, but about how the machine is run: how tightly the system is written down and whether it can bend, and how far ahead it invests.

And then one more, which did not need to be on this list five years ago: how people and machines are meant to work together.

Four about people. Two about the system. One about the thing currently reshaping both.

Nothing in the table above is “more mature” or “better” or “right” – there are just positions that organizations choose to take. Position on who do we treat as talent, how do we build their capability, how do we reward them, move them, how agile our systems are, how long do we plan ahead for and how do we think about intergrating AI.

Why an organization chooses to take a position is crux of reading Talent management systems and a prerequisite to changing it. Not identifying the position. Understanding the why behind the position.

Who counts as talent?

IOCL and McKinsey would both tell you, sincerely, that they take developing people seriously. And you can trust both of them.

IOCL develops everybody. Democratized training that reaches everyone, liberal self-learning reimbursement policy for all, MoUs with renowned institutions, and structured progression pathways. By comparison, McKinsey develops almost nobody – a small identified group receives enormous investment, and a large number of capable people are managed out within a few years of arriving.

Both are right. Work out why before reading on.

IOCL’s output comes from thousands of people executing standardised, safety-critical work reliably, in a business where no individual moves the numbers enough to notice. Its workforce is unionised and its licence to operate is social as much as commercial. Concentrating development on only a chosen few would buy almost nothing operationally and cost a great deal politically.

McKinsey’s revenue sits inside a small number of client relationships held by a small number of people. The gap between its best and its median is genuinely large. Hi-Pos can join a competitor next week and take the clients along with them. Spreading development evenly would starve exactly the people the business runs on.

Same dimension. Opposite settings. Both derived from the business, neither from values.

Amul is the case that shows the setting can be locked rather than chosen. A cooperative built on the output of around 3.6 million milk producers cannot coherently decide that some of them do not count. Exclusion is not available to it, whatever a consultant recommends.

Ask of your own organisation: how many people here have a development plan that is not a form? The answer is the setting, whatever the policy says.

Where does capability comes from?

Tata spends decades building general managers through the Tata Administrative Service. Reliance Jio staffed a national telecom launch almost entirely by hiring people out of other companies. Both are large Indian conglomerates with deep pockets and long horizons. They made opposite decisions.

The question that separates them is not culture. It is two things about the capability itself: how specific it is to your organisation, and how fast you need it.

TAS builds people who understand how the Tata group works — how the holding company relates to operating companies, how a hundred-year-old set of relationships functions. That knowledge does not exist anywhere else, so it cannot be bought at any price. Jio needed telecom operators who had already run a national network. That capability existed, in abundance, inside its competitors. Building it would have taken a decade the launch did not have.

The same question produces two more answers. When the capability is genuinely outside what your organization is for, or the demand is spiky, you borrow it — which is why consulting firms run associate networks and Urban Company runs a partner model rather than employing its service professionals. And when the work is high-volume, rule-bound and encodable, you can now automate the capability rather than employing it at all. So now, apart from your traditional “build, buy, borrow” choices, you have an option to “bot” as well.

That last option is new enough that most organizations have not priced it properly. Automation absorbs precisely the routine work through which junior people used to learn their trade. Remove it and you have not only saved cost – you have removed the bottom rungs of a ladder. I call the result capability debt, and its defining feature is timing: it is cheapest at the moment you decide and most expensive years later, when there is nobody ready for the seat above.

Ask of your own organization: of the last three people who joined the leadership, how many came from inside?

How do people move?

Schneider Electric, an industrial firm of 1,40,000 people, went looking for why its good people were leaving. Close to half of its voluntary turnover turned out to be attributed to a perceived lack of internal opportunity. Not pay. Not managers. People were leaving the company in order to do something different, because they could not see a way to do something different inside it.

Schneider’s response was an open internal market — Open Talent Market — where work is posted as projects and people apply by skill rather than waiting for a vacancy in their function. Unilever built the same thing as FLEX Experiences, rolled it out to more than 100,000 employees across ninety-odd countries, and used it during the pandemic to redeploy around 8,300 people and roughly 300,000 hours of work between businesses that had stalled and businesses that were growing.

Now the opposite, and it is equally correct. Indian Railways moves people vertically, inside a function, through defined grades, slowly. Try running Schneider’s model there and the reason it fails is not cultural resistance. It is that the cost of an inexperienced person in a signalling seat is not measured in money.

That is the real question on this dimension. Is your work project-shaped or stable? And when somebody is under-experienced in a role, does that cost you margin or does it cost you something you cannot buy back?

Ask of your own organisation: name one person who moved from your function to a different one. If the room goes quiet, the setting is closed, whatever the intranet says about internal mobility.

How do you reward talent for their contribution? The interesting thing here is not the quantum or type of rewards, it is what AI is doing to the evidence of who is contributing and how much. Differentiating pay requires being able to say who produced what. As more output becomes joint work between a person and a system, that attribution gets harder – and it dilutes the case for differentiating pay. Notice also that the most aggressively differentiated payer most people can name, Netflix, has no annual performance bonus for regular employees at all. It pays personal top of market in cash and resets whenever a person’s market value moves. Differentiation and incentive schemes are not the same thing, and confusing them is common.

How formal are your talent systems? Codified systems exist to stop an organisation being arbitrary, and they work. However, a system can become so mature that maturity itself is the constraint. Moving even a small part requires significant changes in the system because everything is connected. High maturity turns into exactly the rigidity the process was built to prevent.

Investment horizon. Episodic investment in talent is the most defensible thing in the world at the moment you defend it. Expenditure is mapped to a current problem, the board approves, and there’s no immediate problem. What breaks is the organization’s talent pipeline, and it breaks invisibly until a senior person resigns and there is nobody ready. By the time the cost is legible it is not fixable inside the timeframe anyone needs it fixed.

Humans and machines. The consensus position is that adding AI licences and subscriptions without redesigning roles is a failure of ambition. Sometimes. But holding your job architecture still while you work out where the value of AI actually sits is a legitimate position, not laziness – and it is the correct one more often than the discourse admits. It stops being legitimate when you have known for two years. Part 2 is largely about how to tell those two situations apart.

What to do with this

Read your own organisation on all seven. It takes an afternoon and it will be wrong in places, which is fine. A wrong first reading you can correct beats no reading at all.

Then ask the question that actually produces the insight: which of these settings could your organisation not change, even if it wanted to?

Some are locked by things far outside HR. A cooperative cannot run an exclusive talent lens. A public sector undertaking cannot differentiate pay beyond what the pay scales allow. A safety-critical operator cannot open its internal market to whoever fancies the job. These positions are the business showing through the HR policy – and once you can see which positions are chosen and which are load-bearing, you stop spending credibility on proposals that were never available.

The truth is that you did not choose any of these positions. They were fixed years ago, by people who have since left, in response to conditions that may no longer hold. Most people in HR spend an entire career operating one of the seven without ever learning where the other six are set – running recruitment inside a build-versus-buy decision nobody explained, or performance management inside a reward logic set by a statute they have never read.

The difference between working in talent management and understanding it is knowing all seven positions and how they were arrived at.

Part 2 takes three real organisations – an IT services firm, a dairy cooperative and a quick commerce platform – where the same fashionable intervention was proposed and was the wrong first move in all three, for three completely different reasons. It also deals with what happens when you try to move the organization’s position on one dimension and it lands on another, which is the failure nobody prices in advance.

References

1. Lewis, R.E. & Heckman, R.J. (2006). “Talent Management: A Critical Review.” Human Resource Management Review, 16(2), 139–154. Key insight: Finds the field has no clear definition, scope or conceptual boundaries, with the term used interchangeably for succession planning, high-potential management and HR as a whole. Relevance: The reason to work from a practitioner’s list of decisions rather than waiting for an agreed definition of talent management. Read further: https://www.sciencedirect.com/science/article/abs/pii/S1053482206000271

1. Collings, D.G. & Mellahi, K. (2009). “Strategic Talent Management: A Review and Research Agenda.” Human Resource Management Review, 19(4), 304–313. Key insight: Argues talent management should begin from identifying the positions that disproportionately affect competitive advantage, rather than from identifying individuals. Relevance: The intellectual case for McKinsey’s setting on dimension one, and a corrective to treating inclusion as automatically the more ethical choice. Read further: https://www.sciencedirect.com/science/article/abs/pii/S1053482209000308

1. Lepak, D.P. & Snell, S.A. (1999). “The Human Resource Architecture: Toward a Theory of Human Capital Allocation and Development.” Academy of Management Review, 24(1), 31–48. Key insight: How valuable and how firm-specific a capability is determines whether you should develop it internally, hire it, or contract for it — producing different employment modes inside one organisation. Relevance: The logic that separates Tata’s TAS from Jio’s launch hiring, and the foundation of dimension two. Read further: https://journals.aom.org/doi/10.5465/amr.1999.1580439

1. Cappelli, P. (2008). “Talent Management for the Twenty-First Century.” Harvard Business Review, March 2008. Key insight: Treats talent as a supply chain problem under demand uncertainty; the cost of getting build-versus-buy wrong is asymmetric and routinely underestimated. Relevance: Why over-building is as expensive an error as over-buying, and why the choice is a forecasting problem before it is an HR one. Read further: https://hbr.org/2008/03/talent-management-for-the-twenty-first-century

1. Delery, J.E. & Doty, D.H. (1996). “Modes of Theorizing in Strategic Human Resource Management: Tests of Universalistic, Contingency, and Configurational Performance Predictions.” Academy of Management Journal, 39(4), 802–835. Key insight: Tests three competing views of HR effectiveness against the same data; the configurational view, which treats practices as an interlocking set, adds explanatory power beyond the other two. Relevance: The evidence that these settings must be read together rather than one at a time. Read further: https://journals.aom.org/doi/10.5465/256713

1. Baird, L. & Meshoulam, I. (1988). “Managing Two Fits of Strategic Human Resource Management.” Academy of Management Review, 13(1), 116–128. Key insight: HR practices must fit the business strategy and must also fit each other; either one failing undermines the other. Relevance: The two-part test running underneath all seven dimensions. Read further: https://journals.aom.org/doi/10.5465/amr.1988.4306985

1. Delery, J.E. (1998). “Issues of Fit in Strategic Human Resource Management: Implications for Research.” Human Resource Management Review, 8(3), 289–309. Key insight: Introduces “deadly combinations” — practices that are individually sound and destructive in combination. Relevance: Why reading one setting in isolation is not enough, and the subject of Part 2. Read further: https://www.sciencedirect.com/science/article/abs/pii/S1053482298900066

1. Khilji, S.E. & Wang, X. (2006). “‘Intended’ and ‘Implemented’ HRM: The Missing Linchpin in Strategic Human Resource Management Research.” International Journal of Human Resource Management, 17(7), 1171–1189. Key insight: The gap between HR policy as written and HR policy as experienced explains outcomes better than either does alone. Relevance: Why you read these settings from what an organisation does, never from what its policy manual or careers page claims. Read further: https://www.tandfonline.com/doi/abs/10.1080/09585190600756384