Why a skills review, and why now
Most IT teams know their headcount and their open roles. Far fewer can say which skills they will need in two years, where the team is one resignation away from trouble, or how AI is changing each role. Yet that is what decides whether the strategy can be delivered, and how much of it will depend on partners.
A skills review answers three questions for the IT leadership team: what skills does the team need, at what level; where are the gaps and single points of failure; and what will we do about them: train, hire, partner, or redesign the work. Done well, it takes a few weeks, not a large programme.
Six shifts to build into the review
A skills review that only looks at today's work will plan for yesterday's team. Test each role against six changes already under way:
| Shift | The question to ask in the review |
|---|---|
| AI in every role | What level of AI fluency does each role need, and which of its tasks will AI take over or change? |
| Projects to products | Who will own products and platforms for the long term, with outcomes to deliver, rather than projects to close? |
| Platforms over tickets | Which requests should become self-service, and who has the engineering skill to automate them? |
| Agents with guardrails | Who can manage identity, integration, testing and human oversight for AI that acts across systems? |
| Security and resilience everywhere | Does every engineer build securely by default, and can the team prove recovery, including at suppliers? |
| Every cost visible | Do the people who design and run services understand what their choices cost in cloud and AI usage? |
Use a short scale anyone can apply
Describe each level in behaviour, not years of experience, so that a manager and a team member reach the same answer. Four levels are enough: someone who knows what the skill is and when to call for help; someone who does the work alone on normal cases; someone who handles the hard cases and coaches others; and someone who sets the standard for the organisation. Then write, for each skill, what the second, third and fourth levels look like in practice.
Group skills into a few families that people recognise: deciding what to build and why, building, running, protecting, steering money, work and people, and the human skills. About forty skills in total covers a whole IT function.
The review in six steps
| Step | What you do | Output |
|---|---|---|
| 1. Agree the roles | Start from a common set of role profiles, adapted to your organisation. Map each person to one primary role. | Roles and people mapped |
| 2. Set the target | For each role, the skills that matter and the level needed, for today and for two years ahead. | Role targets |
| 3. Assess | Each person rates themselves; their manager rates them; they agree a level in a short conversation, on evidence: work done, incidents and projects handled, certifications where they matter. Include long-term contractors in key roles. | Agreed levels |
| 4. Calibrate | Managers compare their ratings across teams, so that lenient and harsh raters line up. | Consistent levels |
| 5. Read the results | Gaps by person, coverage skill by skill, single points of failure, readiness for the next role. | Team skills picture |
| 6. Decide | For each important gap: train, hire, partner, or change the work. Name an owner and a date. | Skills plan |
What to look for in the results
- Single points of failure: critical skills held at level 3 or above by only one person. These are your most urgent risks, ahead of any gap.
- Coverage against the two-year target: skills the strategy depends on where nobody reaches the needed level, often data, architecture, product management, security engineering and AI.
- Skills held only by partners: acceptable for commodity work, a risk where the skill is how you differentiate or how you control a partner.
- Roles AI is reshaping: where routine work will shrink, and what those people could move to with the right training.
- Readiness for the next role: who could step into the critical roles, and how soon.
Turn findings into decisions
| Option | When it fits |
|---|---|
| Train | The gap is one level, the person is motivated, and there is time: six to eighteen months |
| Hire | The skill is core to how you differentiate, needed at a high level, and needed soon |
| Partner | The need is temporary, specialised, or a commodity; keep enough skill in-house to manage the partner |
| Change the work | Automation, a platform, or a simpler process removes the need, or changes the level required |
Bring the result to the leadership team on a few slides: the team today, readiness for the strategy, where the team is thin, the largest gaps, AI fluency, and the decisions. Repeat the review every year, and when the strategy or the operating model changes.
Run it with cards on a table
Skills conversations stall on a screen. Printed role cards on a table get people moving, debating what a role really does, and noticing what is missing.
Questions
How often should an IT skills review be done?
Once a year, aligned with the strategy refresh and the training budget, and again after a major change such as a new operating model, a large sourcing decision or a shift in strategy.
Should people assess themselves?
Yes, then agree the level with their manager. Self-assessment alone is biased in both directions; manager assessment alone misses what people do outside their manager's view. The conversation between the two is where most of the value is.
How do we include AI skills?
Treat AI fluency as a skill every role needs at some level, and add the specialist AI skills to the roles that need them: AI engineering, data, evaluation, governance and oversight. Then ask, role by role, which tasks AI will change.
Which IT roles are growing?
Roles around AI engineering and governance, platform engineering, data, security engineering, identity, FinOps and product management are in demand. Which ones you need depends on your strategy and operating model, not on market trends alone.