Before You Automate Anything: The 5-Dimension AI Readiness Test
Editorial Process: Every article begins with my own observations, experience, research or frameworks. I use AI tools deliberately configured around Foresyght’s methods to help me test, structure and refine the work. I verify the claims, make the final editorial decisions and take responsibility for everything published.
Most AI adoption starts the same way.
A tool gets approved. A pilot gets launched. Someone builds a prompt library. A team tries to “automate reporting".
And for a week or two, it looks like progress.
Then the real world shows up.
Outputs vary. Exceptions pile up. Risk questions emerge late (“Can we use client data in this?”). Leadership gets mixed signals — some wins, some mess — and confidence drops.
This isn’t because AI doesn’t work.
It’s because AI is a multiplier. It scales whatever is already true in your business: clarity or confusion, discipline or drift, trust or risk.
So before you automate anything, you need a readiness test.
Not a tech audit. A business readiness test.
Below is a practical 5-dimension framework designed for ops-heavy B2B service firms—where workflows are complex, data is sensitive, and “move fast” must still be responsible.
The Core Principle: Clarity Before Automation
Automation is the end of a sequence, not the start.
If you automate a process that’s unclear, you don’t get efficiency — you get faster inconsistency.
If you apply AI to messy data, you don’t get insight — you get confident noise.
If ownership is unclear, you don’t get scale — you get unmanaged risk.
Readiness is how you prevent that.
The 5-Dimension AI Readiness Test
1) Strategy Readiness
Question: What are we trying to improve — and how will we measure it?
AI projects fail quietly when the objective is vague.
“Improve productivity” is not a goal. It’s a hope.
A readiness-grade objective has four things:
- a clear outcome (what improves),
- a baseline (where you are now),
- a time frame (by when),
- a named owner (who is accountable).
Examples that pass the test:
- “Reduce proposal turnaround from 5 days to 3 days without increasing rework.”
- “Cut client onboarding admin time by 30% while maintaining compliance checks.”
- “Increase delivery consistency by reducing QA defects per project by 20%.”
If you can’t state the goal in one sentence, don’t automate it yet.
Score yourself (0–5):
0 = No clear objective
3 = Objective exists but isn’t measurable
5 = Objective + baseline + owner + timeline
2) Workflow Readiness
Question: Is the work stable enough to automate — or is it mostly tribal knowledge?
AI loves structure. Service firms often run on “it depends".
If a workflow varies wildly across people, accounts, or contexts, AI will not standardise it. It will replicate variability at speed.
Signs you’re ready:
- defined inputs and outputs,
- known steps and handoffs,
- decision rules (what changes the path?),
- a definition of done,
- a QA checkpoint.
A simple test: could a capable new hire follow this process without asking three people?
If not, the best first move isn’t automation. It’s clarification.
Score yourself (0–5):
0 = Entirely ad hoc
3 = Partially documented, inconsistent execution
5 = Documented, repeatable, measurable workflow
3) Data Readiness
Question: Can AI access the right information safely, accurately, and traceably?
In ops-heavy B2B service firms, “data” includes more than databases:
- proposals, SOWs, pricing logic
- delivery artifacts
- playbooks and SOPs
- client communications
- policies and compliance docs
Readiness isn’t “we have lots of data.” It’s:
- Availability: Do we know where the data lives?
- Quality: Is it current and correct?
- Permission: Should AI touch it at all?
- Traceability: Can we explain what informed an output?
If you can’t answer "Where did this come from?” you can’t responsibly scale AI into client-facing work.
Score yourself (0–5):
Get The Civic Reset | Charles Okuonghae’s stories in your inbox
Join Medium for free to get updates from this writer.
0 = Data scattered, unknown permissions
3 = Data accessible but inconsistent/unclear policies
5 = Data mapped, controlled access, traceable sources
4) People Readiness
Question: Who owns outcomes — and how do humans stay in control?
Responsible adoption isn’t giving everyone a tool. It’s building a capability.
At minimum, you need:
- Executive sponsor (sets boundaries and risk posture)
- Operational owner (owns workflow + metrics)
- Steward/governance lead (rules, access, vendor decisions)
- Domain reviewers (quality control and edge cases)
Most failures are ownership failures:
- If nobody owns it, risk grows.
- If everybody owns it, nothing ships.
AI needs a clear operating model: who decides, who reviews, and who is accountable?
Score yourself (0–5):
0 = No ownership, ad hoc usage
3 = Ownership exists but isn’t enforced
5 = Clear roles + review checkpoints + accountability
5) Governance Readiness
Question: Can we move fast without multiplying risk?
Governance doesn’t need to be heavy. But it must be real.
At a minimum, your governance should define:
- what data can/can’t be used (especially client data),
- which tools are approved?
- access levels and permissions,
- when human review is mandatory (client-facing, pricing, legal, HR),
- minimal auditability (what was used, when, by whom),
- what happens when something goes wrong.
A practical stance: Move quickly in low-risk internal workflows. Add review where the work is sensitive, client-facing, regulated, or reputational.
That isn’t red tape. It’s stewardship.
Score yourself (0–5):
0 = No rules, uncontrolled tool usage
3 = Some rules, inconsistent enforcement
5 = Clear policies, enforced boundaries, review + audit path
How to Interpret Your Score
Add your scores across the five dimensions (max 25):
- 0–10: Unsafe to scale
- 11–18: Ready for limited deployment
- 19–25: Ready to scale responsibly
This isn’t about perfection. It’s about sequencing.
Where to Start (If You Want Momentum Without Regret)
In most ops-heavy service firms, the best early wins share four traits:
- stable workflow,
- safe data,
- measurable value,
- low downside risk.
Good starting points:
- internal knowledge retrieval (SOPs, playbooks, standards)
- first-draft creation + mandatory review (proposals, reports, summaries)
- meeting capture → action extraction (with privacy rules)
- QA consistency checks (completeness, formatting, policy compliance steps)
- internal reporting cleanup and narrative (from messy inputs to usable summaries)
Avoid starting with:
- pricing automation without controls,
- client-facing claims without traceability,
- contracts/legal language without review,
- HR decisions without governance.
Let AI earn trust through repeatable wins.
The Final Check: The Multiplier Question
Before you automate anything, ask:
What will AI multiply here?
- clarity or confusion?
- quality or inconsistency?
- Trust or risk?
If you don’t like the answer, you don’t need a new tool.
You need a clearer system.
Take the Free AI Readiness Score
If you want to adopt AI responsibly—and build capability instead of chaos—start with a clear baseline.
Take the Free AI Readiness Score to see where you’re strong, where you’re exposed, and what to prioritise first across strategy, workflow, data, people, and governance.
