The short answer
Start with the tasks and decisions you want to improve. Compare a short list of opportunities against the information, systems and people each needs. Then agree an order, an owner and a checkable result for each piece of work.
Describe the work before choosing a tool
Ask the people doing the job to walk through a recent example. What started it? Which information did they need? Where did they copy, wait, search or correct something? What did the next person receive?
Record the current process in enough detail that somebody else can follow it. A request to make weekly reporting easier is something you can investigate. A request to put AI everywhere does not yet tell you what the business needs.
Compare the opportunities on the same basis
For each task, record its frequency, the manual effort involved, the source of truth, the person who can check the result and the consequences of a mistake. Include the work needed to keep the solution useful afterwards.
These are questions to investigate, not a scoring formula that guarantees a return. A frequent task can still be a poor starting point when its information is unreliable or nobody owns the process.
| Question | What to establish |
|---|---|
| Usefulness | What would improve for the person doing the task, and how will you observe that change? |
| Information | Which source is current, complete and appropriate for the task? Who maintains it? |
| Feasibility | What do the actual systems support? Does the job need training, configuration, ordinary automation or custom software? |
| Review | Who checks the result, what evidence do they see and when must the process stop? |
| Ownership | Who will operate the workflow, update instructions and resolve an exception? |
Worked example: reporting, enquiries or a chatbot?
Imagine a fictional operations team considering three ideas: automate the weekly report, help colleagues answer enquiries and introduce an internal chatbot. The right order depends on the work and its dependencies.
The report has agreed calculations but inconsistent exports. The enquiry team has approved reference material but needs practice checking drafts. The chatbot would depend on a document library that nobody currently maintains. An initial recommendation could be to fix the reporting inputs, train the enquiry team and nominate an information owner before revisiting the chatbot.
| Opportunity | A practical next action in this example |
|---|---|
| Weekly reporting | Agree the export structure, reporting period and calculations. Test a missing column and a repeated file. |
| Customer enquiries | Practise preparing and checking drafts from approved material. Record what reviewers need to correct. |
| Internal chatbot | Assign an owner to the reference material and establish a small, current set of documents before assessing a connection. |
Choose the kind of help the task needs
Some opportunities depend mainly on people knowing how to use a tool: explain the task, provide suitable context and check the answer. Others need a supported connection, a repeatable workflow or an interface that your current software does not provide.
Keep fixed rules for calculations, required fields and permissions. Where an AI step interprets or drafts information, define its input and how a person verifies the output. The strategy should explain these choices in business terms.
Turn the recommendation into an actionable roadmap
For each agreed opportunity, name the owner, the next action, any dependency and the result someone will check. Explain what is included in the work and what still needs investigation.
The roadmap can be short. It needs enough detail to make the next decision possible, including when training or information preparation must happen before implementation.
- Record the current way of doing the task and the result you want to improve.
- Agree which information and systems the work may use.
- Name the people who will test, review and operate it.
- Test ordinary examples, incomplete inputs and cases that should stop.
- Compare the result with the original process, including remaining corrections and manual effort.
- Decide whether to introduce it, adjust the approach or stop.
What should you receive from AI consulting?
A useful consulting engagement should leave you with an explanation of the work reviewed, a prioritised shortlist, reasons for the recommendations and an understandable set of next steps. Agree the depth and deliverables before starting.
Get AI Working offers consulting on its own or alongside training and implementation. You work directly with the founders to assess the opportunities and agree practical next steps. An introductory call establishes fit. Detailed discovery and strategy are separately scoped work.
An AI strategy connects a real business task to an appropriate approach, a responsible owner and a result you can check.
Explore the connected topic
Put the idea to work
AI Consulting & Strategy
You know AI could help. The harder question is where to start. We work through your processes, people and systems to identify worthwhile opportunities and build a practical AI strategy around your business.
How we can help