Skip to main content

AI Strategy

What business problem am I actually solving — and can I say it without using the word "AI"?

The cheese maker knows cheese. They don't need to know Kubernetes. They need a protocol that prevents jargon from replacing clarity and sales pitches from replacing decisions. Same six-box canvasQuestions & Decisions + Plan — but the person filling in the boxes changes everything.

This is a specialized Decision meeting for technology choices. One critical inversion: the business owner chairs, not the tech person. The moment the tech person runs the meeting, it becomes a sales pitch. The protocol gives the owner the questions. They don't need the answers — they need the questions.

Three Roles

RoleJobRule
OwnerChairs, asks questions, decidesStates the business problem. Has final say.
Tech AdvisorPresents options, answers questionsFills the pre-meeting brief. Explains in business language or gets sent back.
Note TakerRecords in decision journal formatCaptures decisions, owners, dates, review cadence.

Red flag: If the Owner IS the tech person, acknowledge the conflict. You can't sell to yourself and buy from yourself in the same meeting. Bring an external advisor or at minimum a skeptical colleague.

Five Phases

Intent

Owner's question: "What business problem am I solving?"

No technology discussion until the problem is named in business language. If the tech advisor starts with "We should implement..." — stop. Rewind to the pain.

  • What process hurts?
  • What does it cost in time, money, or missed opportunity?
  • What does "solved" look like in terms the customer would notice?

Route

Owner's question: "What are my options and what do they cost?"

The tech advisor presents a maximum of three options. Each in a three-column format the owner can read:

OptionWhat it doesWhat it costs ($/month + hours to implement)What happens if we change our mind
A
B
C

"It depends" is not a cost. If the tech advisor can't estimate, they haven't done enough research. Send them back.

Infrastructure

Owner's question: "What do I already have that works?"

Audit before acquisition. Most business owners use 20% of their existing tools. Before buying anything new:

  • What tools do we already pay for? What percentage of their capability do we use?
  • Can the existing tool do 80% of what the new one promises?
  • What's the switching cost — not just money, but workflow disruption and retraining?

Settle

Owner's question: "Who owns this, what's the timeline, what does success look like?"

Every decision gets recorded in the decision journal:

  • Owner: Named person who will make it work (not "the team")
  • Timeline: Implementation date + first review date
  • Success metric: One number the owner can check without asking the tech person
  • Budget: Total commitment including hidden costs (training, migration, downtime)
  • Review date: When we check if this actually solved the problem we named in Intent

Commit with bets — if the tech advisor is confident, they should be willing to stake their recommendation.

Feedback

Owner's question: "Did it solve the problem we named?"

Review meeting at the cadence set in Settle. Compare expected vs actual. Use the same business language from Intent — not technical metrics the owner can't verify.

  • Did the pain described in Intent reduce?
  • Did the success metric from Settle move?
  • What surprised us — good and bad?
  • Continue, adjust, or cut?

Pre-Meeting Brief

The tech advisor fills this in BEFORE the meeting. The meeting is for questions and decisions, not presentations.

SectionContentRule
ProblemIn business language, one paragraphNo technical terms. If the owner can't read it aloud and nod, rewrite.
OptionsThree maximum, each with the Route tableCost in dollars and hours. Not "it depends."
RecommendationOne option with reasoningReasoning must reference the business problem, not technical elegance.
RiskWhat could go wrong, in business terms"Data migration takes longer" not "schema incompatibility."

For quarterly reviews, use the full AI Strategy Review template — it scores current state, gaps, and priorities across 8 sections.

Jargon Gate

Any recommendation that can't be explained in one sentence the owner understands gets sent back. This is the DX validation gate translated for business owners: if the tech advisor can't explain it simply, they don't understand it well enough.

The owner's test: "Can I explain this decision to my business partner over coffee?" If not, the recommendation needs another pass.

Cadence

FrequencyWhenInstrument
MonthlyStandard operating rhythm for most small businessesPre-meeting brief + this protocol
WeeklyDuring active implementation onlyShort check: on track / blocked / need to adjust
QuarterlyStrategic reviewFull AI Strategy Review template

Shadow Risks

Abdication. Owner defers to tech advisor on everything. "You know best" replaces decision-making. The protocol becomes theatre — the tech person decides and the owner rubber-stamps.

Shadow check: Is the owner asking questions or just nodding? If the owner hasn't challenged a single recommendation, they've abdicated.

Analysis paralysis. Owner keeps asking for more options, more research, more time. Nothing gets decided. Set a decision deadline at the start of every meeting. If no decision by the deadline, the default is "no change" — and that's a decision too.

Wrong Meeting?

SignalRedirect to
No business problem identified yetStop. Go find the pain before scheduling a tech meeting.
Team needs to brainstorm possibilitiesCollision — diverge first, then bring options here
Disagreement about what success looks likeDebate — align on criteria before committing
Past decision needs honest reviewAccountability — compare outcomes to expectations
Tech advisor hasn't prepared the briefReschedule. Meetings without prep are parties.

Context

Questions

When the tech person chairs the meeting, whose problem gets solved — the business owner's or the vendor's?

  • If the owner can't state the business problem without using the word "AI", is there actually a problem worth solving?
  • What's the cost of the tools you already pay for but don't use — and is that number larger than the new tool you're considering?
  • How do you tell the difference between a tech advisor who recommends what you need and one who recommends what they know?
  • If every AI decision had a mandatory 90-day review, how many tools would survive the first check?