The 6 Questions I Now Ask Before Any AI Meeting
A framework for navigating AI discussions without wasting time.
It’s Thursday morning as I write this. What a week.
After sitting through two baffling AI meetings in the last few days, I've finally had enough!
One was with a recruitment firm. The other, an education company. Different industries, but identical problems: no one could articulate what they actually wanted AI to do for their business.
This pattern has become so predictable, fueled by the “we need AI in our business“ mantra, that I've developed (out of necessity) a qualification framework – what I’m calling the "AI Compass" for now – to help navigate these discussions more productively.
If you're frequently pulled into similar meetings where enthusiasm exceeds clarity, this might save your sanity too.
I know I’ve already posted twice on SubStack this week, but this I have to share.
Why I’m Doing This
Like I just said, this “compass” has emerged from necessity. From now on, the basics in this template are going out as questions ahead of meetings to:
Qualify whether I'm the right person to help
Ensure the discussion has meaningful direction
Prevent the all-too-common scenario where everyone leaves with nothing but vague excitement
The reality is that AI discussions have become information-harvesting sessions where experts are expected to deliver clarity despite executives not having done the foundational thinking. This “compass” is intended to reverse that dynamic.
The 6-Point AI Compass
Rather than nodding through another"we should do something with AI" session, I've put together a six-stage questionnaire that gives me as much practical insight as I can get before and during meetings so as not to waste time.
I'm sharing this mechanism because I suspect many of you are in exactly the same position I am – trying to bring clarity to AI discussions overwhelmed by hype and uncertainty.
Grab a copy below if you like.
Here's what's in it 👇
1. Where Am I? (Current State Assessment)
Most discussions are starting with, “What can AI do?”. Wrong question.
We really need to be starting here instead:
“Where are we bleeding time, energy, and expertise?”
That’s the real starting point. Because AI isn’t about shiny new toys.
It’s about solving old, expensive problems that we’ve just gotten used to.
Let me give you a real-world pattern I keep seeing.
A mid-sized sales team. They’ve got a CRM full of leads, but most of it is digital dust.
Why? Because account managers are spending hours each week:
Digging up background on companies
Reviewing past emails and notes
Compiling client-specific briefs
Writing intros from scratch
And the important bit about this - None of that requires human judgement. It's research. It’s admin. It’s soul-crushing prep work. And it’s costing you.
So the answer to the “where am I question” isn’t “I need AI” it’s:
“We’ve trapped our best people doing their worst work.”
This is called latent capacity, when talented people are locked in low-leverage tasks.
And the cost of doing nothing? It’s not zero. You could probably measure this in either the number of hours spent each week on the background research, the opportunity cost as in, we’re losing £50k/month in opportunity cost because our team can’t scale outreach or it’s the number of sick days burnt out sales associates and account manager are taking.
This isn’t theory. It’s happening in real companies, right now. And if you're not asking these questions, you're paying for someone else's clarity.
2. Where Do I Want to Go? (Define Outcomes)
If “Where Am I?” exposes the wound then this second stage is intended to set the target for healing. The worst thing you can say here is:
“We just want to experiment with AI.”
Because guess what? Experiments don’t get budget. Outcomes do.
When you define your outcomes, this should be honing in on the signal and cutting out the noise. Let’s go back to our sales team example.
The AI agent we described? It’s not just a time-saver. It’s a capacity multiplier.
If implemented well, it should produce tangible outcomes, like:
2x the number of client dossiers prepared per rep
30% faster onboarding for new account managers
A 15% lift in conversion rate due to better prep
A 10-hour per week reduction in low-leverage admin
Now that’s something you can model, measure, and fund.
There’s another element to think about here too, and that is - what the term ROI means to the leadership team. A lot of the time, a lack of defined outcome means that what they really mean by “what’s the ROI of this?” is “are you sure this isn’t going to lose loads of money“. And in AI, if you’re not clear on your outcomes, that’s very likely where you’ll find yourself.
To avoid that kind of issue, you need to break it down:
What’s the direct financial impact? Will it grow revenue, reduce churn, or cut costs?
Will we reach goals faster or with fewer people? Or both?
Will this give us a competitive edge others haven’t unlocked?
If we can get to answers to these questions quickly, we can establish whether or not AI is the right answer or not.
3. What’s the Terrain? (Know Your Constraints)
You’ve got a clear problem. You’ve defined measurable outcomes.
Now ask:
“What could kill this before it starts?”
Because good strategy isn’t just about what you want. It’s about knowing what’s in the way.
As an agentic solution developer, I cannot overstate the know your constraints part.
Let’s go back to our sales AI agent.
On paper the concept of auto-generate dossiers by connecting an agent to CRM, email, and the web looks great, but in practice:
CRM access is siloed behind team-specific permissions
Email data is governed by strict privacy policies
Web scraping violates TOS of many websites and risks account bans
Your team has no experience integrating APIs or deploying agents
Your AI vendor won’t sign off on data security terms
Now imagine burning £200k getting halfway through before the legal team shuts it down. The risk isn’t really the issue, it’s the surprise and the negative impact of an unacknowledged risk that kills projects and careers.
When it comes to navigating the terrain, this is the man I always think about.
4. Choose Your Path (Strategy First)
This is where you shift from “we should do something with AI” to “this is the kind of change we’re committing to.” And if you skip this step? You’ll waste budget and credibility.
Granted not all AI projects relate to agentics, but since this is what I know, this where I’m best placed to give you a view.
Projects generally fall into one of three buckets:
Automation – Remove work
Augmentation – Support work
Transformation – Rethink work
Each has different expectations, risks, and rewards. The mistake people make is that they reach for transformation, when all they really need is automation. Politics plays a part in this, corporate visibility and career advancement too but the rule generally is, put it in the right bucket.
Let’s revisit the sales agent example and ground this point.
The agent could be put into an automation role to collect client data and email the dossier (saves time, no behaviour change)
It could be setup to prepare dossiers and suggest pitch angles based on previous calls (adds insight, but humans still decide) so as to augment the business process
You could replace manual prep and outreach with a fully autonomous outbound agent (replaces entire workflows, requires trust, and massive buy-in) thus transforming the process
Each path has its cost, its tech stack, its cultural hurdle. Automation is fast, affordable, and safe. Augmentation requires more training and validation.
Transformation is a bet-the-business move where such AI could actually increase enterprise value.
Be clear on which bucket your in.
5. Select Your Vehicle (Pick the Tech)
I’m not going to try and influence anyone in this section, or argue about what’s best because all businesses are different. But I’m a build and not a buy kinda guy.
The truth is the tech doesn't matter until you’ve mapped the mission. Without that, choosing a tool is like buying a sports car to deliver the mail. It looks impressive but it’s expensive and overkill.
Let’s return to our sales example.
You’ve decided to automate dossier prep. You’re keeping it simple. Now the real questions are:
What systems need to connect (CRM, SharePoint, email, LinkedIn)?
What data needs to be extracted, cleaned, and summarised?
Who needs to trust this? (Sales? Compliance? Clients?)
And based on your earlier answers — Are you looking for:
A no-code tool that runs in the background?
An internal agent built on OpenAI or Claude APIs?
Do you want your own model based on something from Hugging Face?
A fully managed SaaS solution with enterprise support?
Each “vehicle” has trade-offs but the real questions to ask in an enterprise context are can it:
Can your IT landscape support it?
Pass your security review?
Be integrated by your current team?
Actually get used by real people next week?
Don’t choose the most powerful tool. Choose the least complex one that gets the job done.
6. Navigate & Measure (Track the Value)
Most AI projects don’t fail because the tech doesn’t work. They fail because the business forgot how to “drive”.
In a talk at the AI Engineering Summit, Stephen Chin dropped a stat:
30% of GenAI projects will be abandoned by end of 2025.
Why?
Not hallucinations. Not model drift. Because they never had a real business use case.
The only fix is to make sure you track, learn and adapt.
A great AI project rollout looks something like this:
Pilot with a small team
Measure usage, time saved, and quality of output
Hold weekly debriefs with users
Pivot if needed. Scale only when proven.
The final point to raise about navigation and measuring with AI is accuracy.
This chart is from an OpenAI team presentation as the same AI summit and for me, perfectly captures the essence of needing to track the value of your project as best you can. You’ll never get to 98% accuracy on the first attempt and planning your rollout in stages because there’s going to be much more than one.
You Don’t Need More AI. You Need More Clarity.
Most execs think the edge is AI capability. It’s not. It’s AI clarity.
Know where you are. Know where you're going. And choose the vehicle based on that.
Enjoy your weekend.
Until the next one,
Chris
🧰 Ready to Stop Being the AI Observer?
The Agent Architect's Toolkit gives you the deep, practical expertise that transforms casual AI interest into career-advancing competence. You shouldn't have to nod along when colleagues discuss agent architecture or watch others position themselves as the AI experts at your company.
Stop being the AI observer. Start being the AI advisor.