Value-Led AI
Most AI saves time.
We use it to create value.
Copilot and ChatGPT make people faster at their own work. Used on the way the company runs, AI takes out cost and lifts margin. We find that value, size it, and build it.
1 to 5 yearsis the window most of our clients are working to.
12%
annual EBITDA growth now needed to hit historic returns. It used to be 5%.
Bain, Global Private Equity Report 2026$3.70
returned for every $1 spent on AI. The best get $10.30 back.
IDC and Microsoft, 2,000 enterprisesOnly 7%
have AI built into how the business actually runs.
FTI Consulting, 2026 Private Equity AI RadarWhy it matters
The price has to be earned.
Between 2010 and 2022, leverage and multiple expansion made up about 59% of private equity returns. Both are largely spent. Thirty-two thousand companies are now waiting to be sold, bought when money was cheap and prices were high.
So the return has to come from the business itself. Bain puts the requirement at roughly 12% annual EBITDA growth to hit historic benchmarks, against the 5% that worked in the last cycle. That is a different job.
AI is the fastest way to move margin right now, if you use it on how the company runs rather than on its inboxes.
Why it is not paying
Almost everyone starts from the tool.
There are two common ways to begin. Both are reasonable. Both stall in the same place.
01
Tool rollout
Buy licences for everyone, run some training, wait.
People get faster at their own work. The document still goes to the same person for the same approval, and still waits three days. Nothing reaches the P&L.
02
Use case hunting
Ask every team where AI could help them, score the answers, build the top ten.
People answer inside their own job, so you get a list shaped like your org chart. You automate the fragmented process you already had, after paying for the build, the integration, the risk work and more people.
The value is not inside the tasks. It is in the handoffs between them, and nobody owns a handoff, so nobody asks for it.
What it costs
<10%
of AI use cases ever leave pilot
Promising projects stall before they reach anything that counts.
Workato1.4x
the cost of a single-use platform, for five times the capacity
Build without shared foundations and every use case starts the cost meter from zero.
phData76%
of enterprise AI use cases were bought, not built, in 2025
Up from 53% the year before. Building your own is losing the argument.
Beam56%
of CEOs saw neither revenue growth nor cost reduction from AI last year
Money spent, nothing in the numbers.
PwC, January 202670%
of one company's AI use case list was the same pattern underneath
Connect to systems, analyse, act. Each one scoped, built and governed separately.
From our own work
What changes
A tool rollout is not an operating model.
Buying a tool changes what your people can do. Changing how you operate changes what the business earns.
AI value
creation
People
- Build an AI-native mindset
- Build capability in the teams you have
- Build new teams where you need them
- Define the new roles and responsibilities
Technology
- Enterprise architecture built for AI
- The AI tool stack, which looks nothing like the last one
- Data security and privacy
- Laid over the systems you already run
Governance
- Who approves what, and risk tiering
- Monitoring and observability
- A record of what the AI did and why
- The EU AI Act and GDPR
Process
- Business processes redesigned around AI
- Not the old process with a tool added
- Work that was queued or re-keyed, removed
- Decisions made where the work happens
Hover a part to see what changes. All four have to change together.
The method
The Value-Led AI Method
Start from the value the business needs, work back to what has to change, and buy the tool last.
01
Weeks 1 to 2
Assess
Where you are now, before anyone designs anything.
- What has already been rolled out, and what came of it
- The structures, tools, systems and data you have today
- What is already in flight, and what it is costing
- Where the governance and decision rights actually sit
02
Weeks 2 to 7
Direction
The stage that decides whether the rest is worth doing.
Ask
Where does the business need to get to, and why is it not there? Not where could AI help, which is the bottom-up question in a suit.
Clean Sheet
If you were designing this service today, knowing what AI can do, how would you run it? Whole journey, end to end, cross-functional people in the room together.
Case
The value cases that fall out of it, sized. The business case. The roadmap. One argument an investment committee can sign, not forty small ones.
03
Runs alongside
Foundations
Build the skeleton once, so use case two costs a fraction of use case one.
- Gap analysis against what Direction says you need
- Buy or build, decided by your size and sector
- Tool stack, data, security, orchestration and monitoring
- The operating model during the change, and after it
04
Three-week cycles
Labs
Ship against the roadmap the business already agreed to.
- One prioritised value case per lab, running by the end of it
- Redesign the process, do not automate the old one
- Change management with the teams whose work changes
- Track the value against the case, through to delivery
Most of the thinking happens in Direction, before anything is built. That is the whole difference.
What we are working on
Where the value came from.
Professional services
Sales capacity
AI took the admin and the lead generation, so the team spend their hours selling.
3x
sales calls from the same team
No extra headcount
Real estate, PE-backed
Legacy ERP replacement
Quoted at £5m over three years. We rebuilt the approach around what the business actually needed.
£4.4m
saved. A £5m quote, tracking to £600K
Three years to nine months
High street retail
The buying function
Discovery across the buying process, systems and team, then a costed case.
£5m
a year, case built
Delivery under way
Who we work with
People who are in a hurry.
Private equity sponsors and their portfolio companies. Owners of businesses planning to sell in the next one to five years. Boards who want the value in the business now and in the price later.
The work takes months, so it has to start while there is still time for it to show. If AI is a line in the five year plan, we are too early for you.
Behind it
Twenty-four years of business transformation.
PwC, Accenture, Wipro and Microsoft. Programmes up to £500m, reporting to the executive board.
Private equity
Building the investment engine
Business transformation for a private equity investor, from operating model through to the systems that run it.
Financial services
$500m programme assurance
Assurance across twelve workstreams of a global payments programme, reporting to the executive board.
Buy and build
Integration planning in a day
A newly acquired business planned into the group in one working session, then delivered.
Multi-site operations
Replacing legacy systems
A legacy estate retired and replaced without stopping the business that ran on it.
The paper
State of Enterprise AI
What it takes for AI to work in a company your size, and to pay back more than it costs. Written for the chief executive setting direction, the finance director signing it off, and the technology leader who has to make it work and keep it safe.
- Why a chatbot rollout stalls
- The seven layers enterprise AI needs
- Where the saving comes from, and where it leaks
- How to prove it small first