Start with a decision
Identify the person using the result and the action it should help them take. A general ambition to use AI is not yet a pilot scope.
- Named workflow
- Useful output
- Measurable review
Six questions worth answering honestly
Most of what decides a first AI project is visible before anyone writes code. Six questions cover it: whether the data you need can actually be reached, whether there is one clear problem to solve, whether leadership is backing the work, whether you have people to build and maintain it, whether the work is repetitive and rule-based, and whether you know your privacy and compliance requirements. A “no” is not a failure. It tells you what to fix first.
- Data you can reach, with a named owner
- One clear problem and a way to measure it
- A sponsor who can make decisions
- People to build it and to keep it running
- An honest view of whether the work is rule-based
- Known privacy and compliance limits
Why first pilots stall
Pilots rarely fail because the model was weak. They stall because the data sits in systems nobody owns, because no one was named to make decisions, because success was never defined so nobody can say whether it worked, or because privacy questions were raised after the build instead of before it. Each of these can be seen in advance.
- Scattered data with no owner
- No named decision-maker
- No definition of success
- Privacy raised too late
Sometimes the right answer is not AI
If the work is predictable and rule-based, conventional automation is usually cheaper and more reliable than a model. If reports cannot be trusted, the first project is data engineering. AI is the right tool when the work involves reading, interpreting or judging: documents, handwriting, images, language. A good readiness review should be willing to tell you that.
- Rule-based work: automate it
- Unreliable reporting: fix the data first
- Reading and interpretation: where AI earns its place
Check the foundations
The readiness self-check covers data, scope, ownership, systems, adoption and privacy. It is a starting point for discussion, not an audit or certification.
- Access to representative data
- A responsible owner
- Privacy and review boundaries
What a readiness report should give you
A useful readiness report is short and specific to your business. It should tell you where you stand today, how others in your industry are adopting AI, the challenges you are likely to meet, and the possible solutions with the benefit of each. Our free report covers these four things and is written by our team from your answers.
- Where you stand today
- How your industry is adopting AI
- The challenges you are likely to face
- Possible solutions and their benefits
Choose the smallest useful next step
When foundations are missing, address those gaps. When they are in place, evaluate a narrowly scoped pilot rather than starting with every workflow.
- Foundation work where needed
- Defined pilot scope
- Evaluation against real tasks
Your next step
When foundations are missing, address those gaps. When they are in place, evaluate a narrowly scoped pilot rather than starting with every workflow.
See how we can help
Build, buy or partner: how to source AI capability
Four routes to AI capability and four questions that choose between them: how central it is, how fast you need it, whether you can hire the talent, and what you want to own at the end.
By Akshay Dodhiwala
Read the article
Running an AI pilot that can actually reach production
Pilots stall when they are run as technology experiments. Run yours as a project: one scope, one owner, a baseline, a small team, weekly checkpoints and a decision to scale, change or stop.
By Ketan Nandani
Read the article