AI workshop or AI assessment: which one do you need?

Start with an AI workshop if your team needs to agree which problem to tackle and where AI might help. Choose an AI assessment if you have a specific task in mind but need to check whether your data, systems and team can support it.

The two approaches can overlap. An assessment may include workshop sessions, so choose by the questions the work will answer and the findings you will receive.

The difference at a glance

Here, an AI workshop means a working session to identify and prioritise possible uses of AI. An assessment checks whether AI could perform a chosen task and what would be needed.

Start with a workshop when priorities are unclear

A workshop is useful when colleagues recognise that work could improve but have different views about where to start.

For example, a service team might spend too much time routing incoming requests. Some colleagues suggest using AI to recommend a category and the team that should handle each request. Others think the bigger problem is that requests arrive without enough information.

A useful workshop brings those colleagues together to review actual requests and follow what happens from arrival to assignment. It helps them distinguish time spent categorising complete requests from delays caused by missing details.

The team might decide to investigate AI suggestions for complete requests and improve the intake questions for incomplete ones. Both address the process, even though only one involves AI.

OpenAI Academy’s use case discovery workshop playbook follows this approach: examine recurring work, identify problems and prioritise applications with owners and next steps.

Use an assessment to check whether AI can support the task

An assessment is more useful when the team has selected a task but lacks the facts needed to decide whether AI can help with it.

Suppose the service team has agreed to investigate AI-assisted request classification. The next questions are more specific:

• Do past requests use categories that still match how teams work?
• Can the application access the information it needs and is that information current?
• Which suggestions need review and who handles incomplete, ambiguous or urgent requests?
• How much work will checking results, correcting errors and maintaining the process add?

An assessment should investigate these questions within an agreed scope. It might establish, for example, that past categories no longer match the current service catalogue. Those categories need reviewing before the records can be used to judge AI suggestions.

Gartner’s guidance for infrastructure and operations leaders recommends assessing use cases for feasibility, risk, cost and expected business impact. The UK Government’s AI Playbook also stresses involving users and maintaining meaningful human control.

The findings should make clear what has been checked, what needs preparing and what still requires testing. That gives the process owner a basis for deciding whether to continue, change the approach or address the gaps first.

When the two approaches overlap

A workshop can reveal questions that require an assessment. An assessment can include workshops to understand the process and agree on responsibilities.

For the request-routing example, a workshop might identify classification as the priority. A focused assessment could then investigate the records, access requirements and review effort involved.

You do not automatically need both as separate pieces of work. If the task and its requirements are already clear, another broad workshop may repeat work. If colleagues still disagree about which problem to solve, resolve that before commissioning a detailed assessment.

Check what your AI workshop or assessment should deliver

Before agreeing to either approach, write down the decision you need help making.

“We want to explore AI” is too broad to define the work. “Which recurring service task should we investigate first?” points towards a workshop. “Can AI categorise requests using the information we have?” points towards an assessment.

Ask what you will receive afterwards. For a workshop, expect a written record of the priorities and follow-up decisions. For an assessment, expect findings that explain what is feasible, what needs preparing and which assumptions remain untested.

For Operations Managers, the decision should include the effect on the whole process. Faster classification may lose its advantage if colleagues spend more time correcting assignments afterwards. Include that work when judging whether using AI for the task is worthwhile.

Our AI Implementation Guide for Business Leaders explores the wider questions around strategy, leadership and implementation.