Does your business need AI, automation or a better process?

Start with the work you want to improve and compare the available approaches. A process change, a feature in your existing software, rules-based automation and AI should all be considered against the same business outcome. The right choice is the one your organisation can justify, operate and improve.

That decision needs more than a demonstration. It needs a clear view of the current process, the cost of the problem and what would have to change for a proposed solution to deliver value.

Define the problem before choosing the technology

Describe the problem in terms a business owner can recognise. “We want to use AI in customer service” leaves too much open. “Our team spends time finding the right policy before it can answer a customer” gives you something to investigate.

Follow the work from request to completion. Identify where people wait, repeat steps, re-enter information or correct errors. Ask which parts need judgement and which follow an established rule. Include the exceptions, because they can determine the cost and practicality of the solution.

Also establish who owns the outcome. A technology team can implement a tool, but the service leader needs to agree what a good customer response looks like and how performance will be judged.

Compare four possible approaches

Use the same problem statement to assess each option:

• Improve the process. Clarify responsibilities, remove an unnecessary approval or make authoritative information easier to find. This may address the problem before new software is needed.
• Use an existing capability. Check what your current systems can do, including features that have not been configured or adopted. Account for the work required to make those features useful.
• Apply rules-based automation. Consider this where inputs, decisions and required actions can be expressed consistently, such as routing a request according to an agreed category or threshold.
• Use AI for a defined part of the work. Consider it where interpreting varied language, drafting or working with unstructured information could help. Decide how its output will be checked and what happens when it is uncertain or wrong.

These approaches can work together. For an illustrative customer enquiry process, ordinary workflow rules could assign the request to a team, AI could prepare a draft using approved guidance, and a person could decide whether the response is appropriate. There is no need to give AI responsibility for every step.

Make the commercial case explicit

Estimate the value of solving the problem before refining the solution. Consider the volume of work, the effort involved, the consequence of mistakes and the importance of a faster or more consistent result.

Then be specific about how value would be realised. Time released from a task creates capacity. A financial benefit depends on what the business does with that capacity, whether costs actually change and whether another bottleneck absorbs the gain. A faster draft may still require substantial checking.

Include the full cost of introducing and operating the approach:

• Configuration, integration and information preparation.
• Licences, usage charges and ongoing support.
• Staff involvement, training and changes to working practices.
• Reviewing outputs, managing exceptions and maintaining controls.
• Rework, future updates and the cost of changing direction.

Treat estimates as assumptions to test. Avoid presenting potential savings as achieved results or treating every minute saved as a reduction in payroll cost.

Check whether the organisation can carry it through

An attractive use case can still be a poor immediate priority. The information may be unreliable, system access may be difficult, or the team may already be carrying a major change programme.

Assess commercial value alongside readiness. A useful shortlist records the problem, expected benefit, dependencies, risks, delivery effort and accountable owner. It should also show what remains unknown and what evidence would change the recommendation.

For example, an internal information assistant may look straightforward until different departments disagree about which procedure is current. Resolving that ownership issue is part of the work. Adding a conversational interface will not settle it.

Choose a first step that produces a decision

Where uncertainty is material, define a bounded evaluation. Use representative work, agree what acceptable quality looks like, and include difficult cases. Measure the complete process, including preparation and review, against the current approach.

Agree the decision before beginning: what evidence would support implementation, further work or stopping? The next step might be a technical trial, a process redesign or an information clean-up. It does not have to be a larger AI rollout.

An effective strategy gives the business a defensible sequence of decisions and clear responsibility for making them. If you need help comparing the options, an AI opportunity review can establish the commercial case, dependencies and a practical path forward.