AI enablement · Readiness and opportunity
Finding the few uses of AI that are worth pursuing
We assess where AI will really pay off in your business, what your data can actually support, and what to buy rather than build. The result is a short list of ideas worth pursuing, and a list of those we recommend dropping.
Where this starts
The problem is never a shortage of ideas
Most organisations we meet have plenty of AI ideas. What they lack is a way to tell the two that would pay off from the twenty that would not. The honest answer usually depends on something nobody has checked: whether the data behind the idea is complete, organised and up to date enough to support it. Enthusiasm gets ahead of that question, and the pilot that follows stalls for reasons that could have been known on the first day.
What usually goes wrong first
- A pilot chosen because it was interesting rather than valuable
- An idea that depends on data whose quality nobody has checked
- A tool bought per user, with no way of knowing whether it changed anything
- Several teams separately building the same thing
- A business case with no starting figures, so the benefit can never be proved
- A custom AI model proposed where a product you already pay for would have worked
Our position
The most valuable conclusion of many of these assessments is that you do not need to build anything. An AI feature in a product you already own will often do the job.
What the work covers
Start with the work, not the technology
We start with the work people actually do, because that is where the time and cost are. Starting with the technology produces a list of things that are possible, rather than a list of things that are worth doing.
Understanding the work and its volume
We measure where the hours go, how much work there is, how often errors happen and what this costs today. Without these starting figures, there is nothing to measure a benefit against later, and the business case becomes a matter of opinion.
Checking the data
We check whether the data behind each idea is complete, up to date and accessible enough to support it. This is the question that most often rules an idea out, and it is far cheaper to ask now than after three months of building.
Build, buy or neither
Most ideas are better served by a product you already pay for. We establish which ones, what they would cost at your scale, and where there is a real gap that needs engineering.
Ranking by value and effort
We give each idea an expected benefit, a range of effort and a level of confidence in both. Your board can then choose between options, rather than simply approve a single recommendation.
Tightly scoped trials
Where the answer is not obvious, we run a limited trial with a measure of success agreed in advance and a fixed end date. Without an end date, a pilot tends to become a permanent cost.
What we advise against
We name the ideas we would not pursue, and explain why. If nothing has been removed from a shortlist, it has not really been assessed.
How the business case stays honest
Starting figures you can be held to
It is easy to claim a benefit before the work starts, and impossible to prove it afterwards, unless you measured the starting point first.
Starting figures
We record current volumes, handling times, error rates and costs before anything changes. The figures come from your systems, rather than being estimated in a workshop.
Measurement
The same measures are built into the process, so the before and after comparison is like for like rather than remembered differently by each side.
Unit cost
We track the cost of each document, ticket or query, including the AI itself. If a benefit disappears at higher volumes, you see it early rather than at renewal time.
Review
Each idea has a review date, agreed at the start, when it is continued, changed or stopped. Stopping is a normal outcome, not an admission of failure.
Register
One place records what is live, what was tried and what was rejected. A second team does not then spend three months rediscovering the same dead end.
What you are left with
- A ranked shortlist with expected benefits and effort ranges
- A data check for each idea, with any blockers named
- A recommendation to build, buy or do nothing for each idea
- Measured starting figures you can be held to afterwards
- A written record of what we advised against, and why
Tell us where your team’s time goes
We will tell you which work is worth automating, which is worth buying a product for, and which is fine exactly as it is.