Illustrative deliverable
What does an AI diagnostic look like?
A worked example of the decisions, assumptions and measurements behind a useful automation plan.
The example business
This is a fictional example, not a client result or a quote. A small service business prepares 40 proposals a month. Each takes 45 minutes of administration before approval. The question is whether reusing approved information and preparing a draft can reduce that work without introducing pricing mistakes.
A diagnostic makes that question testable. It records the current workflow, the proposed change, the assumptions behind the value and the checks needed before wider use.
The opportunity record
| Workflow | Prepare a proposal from an approved brief, rate sheet and standard terms. |
|---|---|
| Current effort | 40 proposals × 45 minutes = 30 hours per month. |
| Proposed change | Draft from approved templates; a person checks scope, price and terms before sending. |
| Target effort to test | 15 minutes per proposal including review = 10 hours per month. |
| Potential capacity recovered | 20 hours per month, if the target is achieved. |
| Illustrative cost assumptions | £30 per hour of staff time, £100 per month of software/support, and £2,000 initial setup. These are example inputs, not SH Applications prices. |
| Main risk | An incorrect scope or price reaches the customer. |
| Control | Use an approved rate sheet, restrict editable fields and require human approval. |
Value, cash savings and payback are different
The illustrative value of recovered time is 20 × £30 = £600 per month. After £100 of recurring costs, the model gives £500 per month of net capacity value. A £2,000 setup cost would have a four-month value-based payback.
That is not automatically £500 of cash saved. If payroll stays the same, the benefit is capacity for other work. A cash-saving claim needs a real reduction in spending; a revenue claim needs evidence that the released time leads to additional profitable work.
If only ten hours are recovered, net capacity value falls to £200 per month and illustrative payback becomes ten months. If running costs exceed the value recovered, the proposal should be changed or stopped.
What happens first?
- Do now: time a representative sample of proposals and record corrections, missing inputs and approval steps.
- Do next: test draft preparation on a small agreed sample. Measure total time, including review and rework.
- Decide: compare actual results with the assumptions and agree whether to expand, revise or stop.
- Leave out for now: automatic sending, free-form price generation and a wider CRM replacement.
How success would be measured
- Median preparation and review time per proposal, compared with the baseline.
- Number of corrections and any pricing or scope errors.
- Total monthly running and support costs.
- How the recovered time was actually used.
A real diagnostic adapts this record to your business and identifies an owner for each next step. Read the valuation method or explore AI Diagnostic & Delivery.
Start with a workflow from your business
Describe the task, how often it happens and where it gets stuck. We can discuss what would need to be measured before making an investment decision.