Before diving into the creation of custom agents, this framework helps you find the right problems and decide which ones to tackle first. We are happy to support you through this process as required.
It works in three steps:
Discover — surface the tasks worth considering
Score — rate each on frequency and time cost
Prioritise — focus on the highest-impact candidates
The output is a shortlist of use cases ready to take into the prompt library. We can help you with the setup or refinement as required.
Step 1 — Discover: Find Your Candidates
Good AI automation candidates are usually hiding in plain sight. Work through the questions below for each team or role in your business. Note down every task that comes up.
Discovery Questions
Volume and repetition
What tasks does your team do more than once a day? More than once a week?
What do you find yourself copy-pasting regularly?
What requests come into your inbox on repeat (from colleagues, customers, or suppliers)?
Time and frustration
What takes longer than it should?
What do people dread doing?
What would your team celebrate if it disappeared from their to-do list?
Knowledge and lookup
What questions do people ask you (or Slack/email) that you've answered a dozen times before?
Where does your team spend time hunting for information that should be easy to find?
What onboarding knowledge exists only in people's heads?
Documents and communication
What documents do you create that follow a similar structure every time?
What customer or supplier communications follow a template but still require manual effort?
What reports do you produce manually from data that already exists in a system?
Capability mapping to your core use cases
Use these to prime your thinking for your business. Later with custom connectors available this scope will continue to expand.
Step 2 — Score: Frequency × Time Cost
For each candidate task you've identified, assign two scores:
Frequency — how often is this task performed?
Time Cost — how long does each instance take?
Impact Score = Frequency × Time Cost (max 25)
Use a simple table to capture this:
Step 3 — Prioritise: The Automation Sweet Spot
Plot your scored tasks against this matrix. Anything in the top-right is where you start.
Where to focus:
Automation Sweet Spot (high frequency, high time cost) — biggest ROI, start here
Volume Reducers (high frequency, low time cost) — fast to configure, adds up quickly
Handle Manually (low frequency, high time cost) — worth doing, but not urgent
Deprioritise — skip for now
Step. 4: Is It AI Agent-Ready?
Once you have your list of top impact items now run each top-scoring task through a quick check:
If this top priority opportunities pass the AI-readiness check then they are your priority candidates. Our Tai consultants would be happy to guide and support on this step.
Step 4 — Your Shortlist
Aim for 3–5 use cases to start with.
For each shortlisted use case, capture:
Name — a clear, plain-English label (e.g. "Draft pre-departure customer email")
Who does it — the role or team this helps
Current process — how it's done today (tools, steps, inputs)
Desired output — what a good Agent response looks like
Success measure — how you'll know the Agent is adding value
This becomes your input to the use case library, where you'll find ready-made prompts or build your own.
Example: Walking Through the Framework
Company: A mid-size tour operator with a team of 8 travel consultants
Discovery finding: Consultants spend 20–30 minutes each time writing personalised itinerary descriptions for proposals. They do this 3–4 times per week each.
Score:
Frequency: 4 (several times per week)
Time cost: 4 (20–30 mins)
Impact score: 16 — lands firmly in the Automation Sweet Spot
AI Agent-readiness check: Clear output (written description), structured inputs (destination, stops, duration, customer type), low risk (consultant reviews before sending) → passes
Shortlisted use case: "Generate personalised itinerary description for customer proposal"
→ This goes straight into the use case library to create an custom agent.