"Organisations that invest in structured AI discovery and assessment before implementation are 2.3× more likely to achieve measurable ROI within 12 months, compared to those who begin with technology selection." — McKinsey Global Institute, February 2026
Most businesses approach AI backwards.
They hear about a tool, or a competitor does something interesting, or they attend a webinar — and they start there. Technology first, problem second. The result is an AI implementation in search of a use case, and that's usually how you get expensive pilots that go nowhere.
The better approach: find where AI will make the most difference in your specific business before you touch any technology. That's what a discovery process does, and it doesn't have to cost anything.
Here's a framework you can run yourself in a focused day.
Step 1: Map Your Most Expensive Processes (90 minutes)
Gather the people who understand how your business actually operates day-to-day. You want the operators, not just the executives.
Ask one question: "Where does time go that shouldn't?"
Document every process that someone identifies as:
- Repetitive (done the same way repeatedly)
- Time-consuming (takes hours per week or more)
- Error-prone (mistakes happen and cost something to fix)
- Bottlenecked (slows down something else when it's slow)
Don't evaluate yet. Just list. You want 15–30 items.
For each item, capture: who does it, how often, approximately how many hours per week, and a rough estimate of the cost of errors or delays.
Step 2: Score for AI Suitability (60 minutes)
Not every repetitive task is a good AI candidate. Rate each item on your list across four dimensions (1–3 each):
Data availability: Is the information involved digital and accessible? (3 = yes, structured; 1 = paper-based or scattered) Consistency: Does the task follow a pattern, or is every instance highly unique? (3 = highly consistent; 1 = totally bespoke) Volume: How many times does this happen per week or month? (3 = high volume; 1 = rare) Impact: How significant is the outcome of this task to revenue or operations? (3 = high impact; 1 = minor)Add the scores. Tasks scoring 10–12 are your primary targets. 7–9 are secondary. Below 7, deprioritize for now.
Step 3: Validate the Top 3–5 Candidates (60 minutes)
Take your highest-scoring items and do a quick feasibility check for each:
Can you describe the task as a clear input and output? If a human sat down to do this task, what would they start with and what would they produce? AI works best when this is clear. Is there a human who currently does it well? That person's process is your training set. If you can describe what they do, you can often approximate it with AI. What would you accept as "good enough"? AI rarely achieves 100% accuracy. If 85% correct with human review is acceptable, many things become viable. If it has to be perfect every time, your requirements are tighter. What happens when it goes wrong? Map the failure mode. If a mistake causes a customer complaint, that's manageable. If it causes a compliance issue or financial error, your risk tolerance is lower and your governance requirements are higher.Step 4: Estimate the Value (30 minutes)
For each candidate, do a rough annual value calculation:
(Hours saved per week × 52 × Loaded hourly cost) + (Errors reduced × Average error cost)You don't need precision — you need order of magnitude. Is this a $10K/year opportunity or a $200K/year one? That shapes how much you should invest in the solution.
The highest-value items with the highest suitability scores are your starting point.
Step 5: Define Your First Project (30 minutes)
Pick one. The smallest version of the highest-value opportunity you found.
Write a one-paragraph brief: What is the task? What's the input? What's the expected output? What does success look like at 90 days? Who internally will own it?
That brief is what you take into a vendor conversation, an AI consultant engagement, or an internal build conversation. It focuses the work and makes evaluation easy.
What Discovery Surfaces That You'd Miss Otherwise
Done well, this process often uncovers opportunities that aren't obvious from the outside:
- A customer communications workflow that's taking 20 hours a week to personalize manually
- An internal reporting process where one person spends two days a month pulling data together
- A proposal or quotation process where most of the content is templated but treated as custom
- An onboarding workflow that every new hire goes through but that's never been systematized
These aren't glamorous. But they're where AI delivers clear, measurable, defensible ROI — and where you build the organizational trust to take on bigger projects.
The Free Part
The discovery process outlined here costs nothing except the time of your team for a focused day. It doesn't require AI tools, consultants, or technical expertise.
If you want external perspective — someone who's seen this process across dozens of businesses and can pressure-test your findings — that's where a short consulting engagement makes sense. But the initial discovery is something every business can do.
Start there. Know what you're solving before you decide how to solve it.
Want to implement AI? Book a call at blok-ai.com