AI in Process
Move from the AI playground to a considered way of working. Understand how your business is using AI, put practical policies and processes in place, and decide which tools genuinely add value to which tasks.
THE CHALLENGE
People are trying prompts, signing up to tools and finding their own shortcuts. That curiosity can be useful, but the business may not know what information is being shared, how outputs are checked or whether the work is actually improving. AI in Process brings those experiments into the open and creates a clear, usable framework for everyday decisions.
WHO IT’S FOR
For owners, leadership teams and operations managers whose people are already experimenting with AI, but who need visibility, reliable advice and a consistent approach before informal habits become business practice.
WHAT WE LOOK AT
Map the tools, embedded AI features, tasks and information flows across the business. Speak to the people doing the work, including those using personal accounts or informal workarounds. Create a useful starting picture without treating every experiment as a problem.
Set out approved uses, restricted information, account requirements, responsibilities and escalation routes in language your team can use. Make it clear what people can do, what needs approval and when to stop and ask.
Turn policy into everyday steps: preparing information, using the tool, verifying the result and approving work before it reaches a client or informs a decision. Define who checks what and what happens when the output is wrong.
Assess fit, quality, data handling, cost and the consequences of failure. Compare a focused trial with the current way of working, including the time spent checking and correcting results. Keep decisions grounded in business value.
Help teams distinguish a plausible answer from a supported one. Establish source checks, current and appropriate inputs, and clear limits on relying on generated content. Translate tool capabilities and limitations into practical guidance for your work.
Agree who maintains the tool register, approves new uses and keeps guidance current. Give people relevant examples and a route for questions. Review changes in tools and working practices with your IT, security or other specialist advisers where needed.
THE AUTOMATE VS PROTECT MATRIX
An AI in Practice framework for discussing
opportunity, risk and human responsibility.
Consider stable, predictable tasks with defined inputs and exception handling. Start small and check that the workflow behaves as intended.
Use AI where it can help draft, organise or interpret information, with a person able to check and correct the output.
Retain meaningful review where decisions affect people, relationships or important business outcomes.
Give sensitive information and high-consequence work additional care. Restrict or pause a use where appropriate safeguards are not in place.
These are discussion categories, not automatic scores or a compliance test. A process may need more than one approach.
FROM THE AI PLAYGROUND TO PRACTICAL DECISIONS
Before another subscription or rollout, use a consistent set of questions. The result should be a decision people can explain and revisit.
Name the problem, the intended user and the result you need. Is the work predictable, judgement-led or sensitive? Could an existing feature, simpler automation or process change do it better?
Identify what goes into the tool and where it goes. Check permissions, confidentiality, account settings and supplier terms before using business or client information.
Agree what good looks like, which sources need verification and who can spot errors. If no one can meaningfully check the result, reconsider the proposed use.
Trial on appropriate examples and compare quality, time, checking effort and total cost with the current process. Set a review point before wider use.
Record the decision: approve for a defined use, trial with safeguards, request specialist review or do not use. Agree an owner and a review date.
HOW AN ENGAGEMENT COULD WORK
Agree the scope, speak to the team and map current tools, tasks and information flows. Identify useful experiments as well as gaps in oversight.
Prioritise the issues and develop proportionate policies, task-and-tool decision criteria and workflow checks. Agree responsibilities with the people who will use them.
Walk the team through relevant examples, agree the first actions and set review dates. Refine guidance as people put it into practice.
WHAT YOU CAN WORK TOWARDS
We agree the scope and intended outcomes together. What changes depends on your starting point and the work you choose to take forward.
QUESTIONS, ANSWERED
Yes. This work connects business operations, governance, AI strategy and technology. Your IT provider, security specialists and internal teams remain important partners. Their technical knowledge helps turn business requirements into appropriate controls.
The aim is to give useful experimentation clear boundaries. A small trial with appropriate information, an owner and a review point can help you learn without treating every new tool as ready for everyday business use.
The aim is to give useful experimentation clear boundaries. A small trial with appropriate information, an owner and a review point can help you learn without treating every new tool as ready for everyday business use.
It may be. Start by establishing what is actually happening. Staff may be experimenting independently, and existing software can include AI functionality. The review is about understanding your situation, not assuming there is a problem everywhere.
A policy is one part of a wider approach. It does not guarantee compliance. Relevant obligations depend on the business and its activities, and legal, data protection or security advice may be needed from the appropriate specialists.
Yes. The scope can concentrate on a team, workflow or set of tools where you most need clarity. We agree the boundaries and expected outputs before work begins.
A SENSIBLE NEXT STEP
Start with the problem you are experiencing. We can work out what kind of help would be useful.