AI workshop, AI assessment or straight to implementation?

You want to do something with AI, and four kinds of engagement are on the table. A workshop to bring the team along. A scan for a first inventory. An audit that costs out processes and picks an order. Or building straight away. They are often compared on price, but that is the wrong comparison. They deliver different things.

Choose by the decision you need to be able to make afterwards. That is the thread running through this article.

Four routes, four kinds of outcome

AI workshop: understand

A workshop gets the team speaking the same language and shows what AI can and cannot do. The outcome is a shared picture, a list of ideas and support. Suitable when expectations in the management team differ, when employees are uncertain about AI, or when nobody has tried it on their own work yet.

Not enough when you need budget approval for an implementation, when several processes need a reasoned comparison, or when integrations and risks need assessing.

AI scan: explore

A scan or quick scan is a first, limited inventory. The outcome is a shortlist, a few initial risks and advice on whether a closer look makes sense. Suitable for a small scope, a single department, or when you want an outside view before investing more.

Note: the term is not standardised. Ask in advance how many processes, which interviews, which data and what outcome are included. One scan is a questionnaire, another is half an analysis.

AI Opportunity Audit: decide

An audit analyses processes, compares opportunities on value, effort and risk, and puts them in order. The outcome is a decision: which process first, roughly what it costs, what it can deliver and what is needed. Suitable when several departments have ideas, when management needs a rationale to allocate budget, or when you do not yet know which technology or partner fits.

Pilot or implementation: build and measure

Building straight away fits when the process, data, owner and success criterion are already clear. The outcome is a working application with measurements and a decision on scaling up. Start here too early and you build something only to find the wrong process was chosen. Why that happens so often is covered in why AI pilots stall.

The differences side by side

• Main goal: workshop to learn, scan to explore, audit to decide, pilot to build and measure.
• Use of process data: workshop limited, scan high-level, audit extensive, pilot essential.
• Business case: workshop usually not, scan indicative, audit per opportunity, pilot actually measured.
• Integrations: workshop as examples, scan in outline, audit on feasibility, pilot built and tested.
• Outcome: workshop ideas and support, scan a shortlist, audit a reasoned order, pilot a working application.

The question that makes the choice

Do not ask which route is cheapest. Ask which decision you need to be able to make responsibly once it is done.

• "We want the management team to understand what AI can do for us." Choose a workshop.
• "We want to know whether there are enough opportunities at all." Choose a scan.
• "We have twenty ideas and need to decide what comes first." Choose an audit.
• "We know classifying incoming email is our first process, and we have the data and an owner." Start a pilot.

A decision tree in six questions

Does the team have enough basic knowledge? No: start with a workshop. Yes: go to the next question.

Has one clear process been chosen? No: do several processes need a reasoned comparison? Only a first orientation: scan. With value, risk and order: audit.

Are the volume, data and process owner of the chosen process known? No: do a process analysis or audit first.

Have integrations, permissions and human oversight been defined? No: technical design first. Yes: start a limited pilot.

Are the pilot results stable? No: improve or stop. Yes: roll out in a controlled way and scale up.

When a workshop is the right choice

A workshop is valuable when terms get mixed up, when teams have different expectations, when usage rules are missing or when management wants support first. It works best if you agree in advance which processes serve as examples, which questions need answering, who owns the next step and when the next decision is due.

A workshop without an owner and a follow-up date stays stuck at inspiration. What our workshop for a management team involves is on the AI workshop page.

When you no longer need a workshop

Many companies are already past the introduction stage. If you recognise three or more of these signs, you do not need inspiration but a choice.

• Your team already uses AI tools, but without coherence.
• There are several ideas and nobody knows where to start.
• Capacity is under pressure, but there is no overview of where most time goes.
• You want to free up budget and lack a rationale.
• There has already been a workshop and nothing went live afterwards.

When an audit delivers more

A deeper analysis makes sense when departments have different priorities, when systems and dependencies are complex, when security and privacy weigh heavily, or when several solutions or suppliers are possible. An audit is not a goal in itself. It should end in a decision on the first application you will actually build. How we do that is on the AI Opportunity Audit page.

When building straight away is sensible

You can start right away if you can answer these questions with data:

• Which process, and which problem are we solving?
• What is the volume now, and how long does it take now?
• Which data and which systems are needed?
• Who owns it, and who will run the solution?
• What may AI do and not do, and where is the human check?
• When is the pilot a success, and when do we stop?

If several answers are missing, analysis comes first. The approach to the build itself is on the AI implementation page.

Three wrong choices that come up often

Workshop after workshop. Lots of ideas, no owner, no process data, nothing live. Stop the broad inspiration, choose processes, measure how things work now and make a decision.

Building too early. The tool has been chosen while the problem is still broad and nobody knows who will run it. Step back to process analysis and design.

Too much analysis. The process is clear, the data and owner are there, the risk is limited, and still months of research are proposed. Start a small pilot with stop criteria.

Combining works, if the workshop stays short

A workshop and an audit do not exclude each other. A short session with the management team to align language and expectations, followed by an analysis of the processes and a decision on the first application, works well. The workshop then speeds things up rather than being the end point. Agree in advance what step two is.

Compare on more than price

Include in your comparison: the internal hours it takes, how many processes are covered, whether a business case comes out of it, whether risks and integrations are assessed, and whether the outcome can be handed to whoever builds it. A cheap engagement that produces no decision becomes expensive through delay. A heavy analysis for a simple process is just as wasteful. How to cost out a process is in calculating the AI business case. What a solution costs is on the cost of an AI solution.

Next step

Not sure yet whether you mainly need knowledge, a choice or delivery? Walk through the decision tree above. If you land on "compare several processes", look at the AI Opportunity Audit. If you land on "the team needs to come along first", the AI workshop is the logical start.