I run a lot of these calls, and an AI automation discovery call checklist can make the difference between a useful thirty minutes and a polite waste of time. It all comes down to what the founder brings into the room, not slides or a polished brief, but a handful of specific details that help us skip the guessing and get straight to the real problem.
Most people arrive with a vague sense that AI should be doing something in their business, but they struggle to explain what that should be. As a result, much of the call is spent excavating information instead of diagnosing the opportunity. Bringing six essential details gives us more time to assess the situation, identify practical possibilities, and determine whether automation is genuinely worth pursuing.
Here is exactly what to bring, what to leave at home, and which questions you should be asking me during the call.
Key Takeaways
- An AI automation discovery call checklist should cover your numbers, your tool stack, your week, your bottleneck, your team, and your honest budget range.
- Bring one week of real time tracking, not an estimate. Founders underestimate admin time by roughly half when guessing from memory.
- List every tool you pay for and whether it talks to anything else. Disconnected tools, not missing tools, are usually the actual problem.
- Name the one task that only you can do. That single answer shapes more of the recommendation than anything else on the call.
- Skip the solution shopping. Arriving asking for “a chatbot” narrows the diagnosis before anyone has looked at the problem.
- Ask what happens if it fails, who owns the build, and what the first thirty days look like. Vague answers there are the reddest flag you will get.
- A good discovery call ends with a clear yes, no, or not yet. If nobody will say no to you, you are on a sales call, not a discovery call.
Why Most Discovery Calls Go Nowhere
The default version of this conversation is a founder describing symptoms and a vendor describing products, and the two never quite meet. You say you are drowning in admin. They say they do AI automation. Thirty minutes later, everyone agrees to “send some information over” and nothing happens.
That failure is not a personality mismatch. It is a data problem. I cannot recommend anything useful if I do not know how many hours your week actually contains, what systems already hold your data, and what breaks when you go away for a week.
The research backs the pattern up. An MIT study of enterprise AI projects found that roughly 95% deliver no measurable return, and the ones that succeed start with a process, not a tool. The same thing happens at small business scale. People buy the software first and work out the use case afterwards. The discovery call is where you avoid that, but only if you arrive with the raw material.
There is a second reason it matters. A discovery call is a two-way audition. You are working out whether the person on the other end understands operations or just knows how to say “agentic” with a straight face. Bringing specifics is how you test them. Vague inputs get vague answers from everyone, so you learn nothing about who is actually good.
The AI Automation Discovery Call Checklist: Six Things to Bring
Here is the list. It takes about forty minutes to pull together, and it will change the quality of every conversation you have with any provider, not just me.
1. One week of real-time tracking
Not an estimate. A week of actual entries. Every time you switch tasks, write down what you moved to and roughly how long you spent. A notes app is fine.
I ask for this because memory lies. Founders consistently guess low on admin and high on strategic work. When someone actually tracks a week, the invoicing, the chasing, the quoting, and the “quick questions” from the team usually add up to somewhere between twelve and twenty hours. That number is the whole business case. Without it we are both guessing.
If you only have two days, bring two days. Real data on two days beats a hunch on five.
2. Every tool you pay for, and what talks to what
Write a list. CRM, accounting, project management, booking system, email platform, phone system, the spreadsheet that one person maintains. Next to each one, note two things: what data lives in it, and which other tools it connects to.
Most of the time the answer to the second question is “none of them.” That is the finding. Founders assume they need more software when the actual gap is that their existing systems do not share anything. Your accounting system knows what a customer is worth. Your CRM knows who they are. Neither one tells the other, so nobody can act on both.
This list also stops the conversation from drifting into a platform migration you do not need. When I can see the stack, I can usually build around it rather than replace it.
3. The one task only you can do
Answer this before the call, honestly: what is the single thing that stops you from taking two weeks off?
For a builder, it is usually quoting. For a broker it is compliance sign-off. For a clinic owner it is anything involving a difficult patient. For an agency owner, it is the client relationship that never got handed over.
This one answer shapes more of the recommendation than anything else. It tells me where the key-person dependency sits, which is almost always a different place from where the founder thinks the pain is. People say “I need help with admin” and then describe a bottleneck that is entirely about decisions nobody else has the context to make.
4. Your lead and enquiry numbers
Three figures. How many new enquiries you get a month, roughly how long it takes before someone responds, and what a closed customer is worth on average.
You do not need these to the decimal. Ballpark is fine. But without them, nobody can size the opportunity, and any provider who quotes you a return without them is making it up.
Response speed matters more than most people expect. Harvard Business Review’s research on lead response time found that firms responding within an hour were around seven times more likely to have a meaningful conversation with a decision maker than those waiting even two hours. Most small businesses I speak to are measuring their response in hours, sometimes days, and have never connected that to the deals they lose.
5. Your team, and what each person actually holds
List every person, including part-timers and contractors. Beside each name, write what only they know.
If one name appears more than about five times, you have found a single point of failure. Usually that name is yours. Occasionally it is a long-serving office manager who holds the entire operational memory of the business and is three years from retirement.
This matters for the recommendation because automation and hiring solve different problems, and the wrong one makes things worse. Adding people to a business with no shared context spreads the chaos rather than absorbing it. Knowing where the knowledge actually sits tells us which comes first.
6. An honest budget range
Not a number you are committing to. A range you would not be embarrassed by.
I know this feels like handing over your hand before the negotiation. It is not, and here is why. If your range is far below what the work costs, I would rather tell you in minute ten than in week three. If it is well above, you should be looking at a bigger scope than the one you came in asking for. Either way, saying it early saves you a fortnight of proposal ping-pong.
If you genuinely have no idea, say that. “I do not know what this costs; help me calibrate” is a completely legitimate answer, and comparing against a part-time hire is usually the fastest way to get there. Our guide to AI automation pricing covers how the numbers typically break down.
What to Leave at Home
Three things that consistently derail these calls.
A pre-chosen solution. Arriving with “I want a chatbot” or “I need an AI receptionist” collapses the diagnosis before it starts. You may be right. You are also constraining the conversation to one answer, and the thing costing you the most money is often somewhere else entirely. Bring the symptom, not the prescription.
The full org chart and a strategic plan. You do not need to prepare a document. Nobody is grading your presentation. The checklist above is deliberately made of facts you already have, not analysis you have to produce. If preparing for a discovery call takes you a full day, the preparation has become the procrastination.
The performance of being fine. This is the big one. Most founders open these calls with a version of the business that sounds more organised than it is. I understand the instinct. It is your name on the door. But if you tell me the follow-up is handled and it is actually handled about a third of the time, we will design around a problem you do not have and miss the one you do.
The whole value of the call is an outside view of the actual operation. You get that by being straight about what is held together with tape.
Questions to Put on Your Side of the Checklist
An AI automation discovery call checklist works in both directions. You are assessing them as much as they are assessing you. Six questions worth asking.
“What would you tell me not to automate?” Anyone who says everything is automatable is selling. There are tasks where the judgment call is the whole job, and a good provider names them without being pushed.
“Who actually does the build?” Ask whether it is the person on the call, an internal team, or a subcontractor you will never meet. Ask what happens if that person leaves. This is where a lot of small providers fall apart.
“What does the first thirty days look like, specifically?” You want dates and deliverables, not phases. Vagueness here predicts vagueness later.
“What happens when it gets something wrong?” Every automated system will handle an edge case badly at some point. The question is whether there is a human review step, an escalation path, and someone who notices. Systems built without a human in the loop fail quietly, which is the worst way to fail.
“Do I own it if we part ways?” Where does the data live, who holds the accounts, and what can you take with you. Worth settling early. Our AI automation contract checklist goes deeper on the terms that matter.
“What would make you tell me this is not a fit?” The best answer to this question is a real one. I have told people to go and fix their pricing first, or hire an admin before automating anything, because the system would have amplified an underlying mess. If nobody is willing to disqualify you, you are not on a discovery call.
What a Good Discovery Call Actually Produces
Thirty minutes is not long enough to design a system. It is long enough to do three things properly.
First, name the bottleneck. Not the symptom you arrived with, the constraint underneath it. Often those are different. Someone comes in wanting faster lead follow-up and leaves realising the real problem is that they are the only person who can quote, so faster leads would just create a longer queue at their own desk.
Second, size it. Rough numbers on what the bottleneck costs per month in lost deals, wasted hours, or delayed work. This is why the enquiry figures and the time tracking matter. Without them, there is no maths, and without maths, there is no decision.
Third, a clear next step, including “nothing.” A good call ends with a yes, a no, or a not yet, and all three are valid outcomes.
What it should not produce is a proposal on the spot. If someone can price your build after thirty minutes with no look at your systems, they are selling a template. The honest sequence is a conversation, then a proper look under the bonnet, then a scope.
I have watched what happens when the diagnosis is right. One dental practice I worked with was losing close to half their inbound calls to voicemail and busy signals, with two receptionists already on the desk. That was never a staffing problem. Once the phone was handled properly, bookings went up 44%. The fix was cheap. Finding the right thing to fix was the work, and that started on a call exactly like this one.
The Bigger Picture
Every item on this list points to the same underlying problem: your business runs on knowledge trapped in your head, systems that refuse to talk to each other, and daily operational decisions that bottleneck through a single person. Individual automations provide temporary relief, but what truly transforms the shape of your workweek is building a centralised intelligence system that connects your data, maintains context, and deploys digital workers to handle repetitive tasks.
You do not need to resolve every operational issue during a single session, but you do need an accurate starting point. Using an AI automation discovery call checklist gives you that clarity before the meeting starts. When you arrive with your weekly schedule, tech stack, key metrics, primary bottlenecks, team structure, and an honest budget range, thirty minutes is plenty of time to determine whether an automation build makes sense.
If nothing else, commit to completing the initial time tracking exercise. Even if you choose never to speak with an automation specialist, documenting a week of daily tasks will reveal more actionable insights about your operations than most standard consulting audits.
Want to run through it? I do a 30-minute Discovery Call where we work through exactly this. No pitch deck, no obligation, and I will tell you if it is not a fit. Book a time here. Bring the checklist.
Frequently Asked Questions
What is an AI automation discovery call?
It is a short, structured conversation to work out whether AI automation makes sense for your business and where it would go first. A good one is diagnostic rather than promotional. The provider asks about your processes, your systems, and your numbers, then tells you what they found. It should end with a clear recommendation, including the possibility that you should not proceed yet.
How long should a discovery call take?
Thirty minutes is enough for a first conversation if you arrive prepared. Longer than an hour usually means either the provider is presenting rather than listening, or the scope is genuinely complex and needs a proper paid assessment. If someone wants ninety minutes for a first call with no agenda, ask what they plan to cover.
What should I prepare before an AI automation consultation?
Six things: one week of real-time tracking, a list of every tool you pay for and what connects to what, the one task only you can do, your monthly enquiry volume and average deal value, a list of your team and what each person uniquely knows, and an honest budget range. It takes about forty minutes to assemble, and it changes the entire quality of the call.
Do I need to know what I want automated first?
No, and it is often better if you do not. Arriving with a fixed solution narrows the diagnosis before anyone has looked at the problem. Bring the symptoms, the constraints, and the numbers. Working out what to automate first, and in what order, is the job of the call. A provider who cannot do that without being told is not adding much.
What questions should I ask an AI automation provider?
Ask what they would tell you not to automate, who actually does the build, what the first thirty days look like with dates attached, what happens when the system gets something wrong, whether you own the assets if you part ways, and what would make them say you are not a fit. Their willingness to disqualify you is the strongest signal you will get.
Is a discovery call the same as a sales call?
They overlap, and pretending otherwise is silly. The difference is direction. A sales call moves toward a close regardless of what surfaces. A discovery call can legitimately end in u0022not yetu0022 or u0022go fix this other thing first.u0022 If every answer you give leads back to the same product, you are on a sales call.
About Octavius
Titus Mulquiney is the founder of Octavius AI, where he builds AI brains and AI workforces for founder-led businesses stuck running everything out of their own head. Twenty years in marketing, ex-Sony product manager, ex-GM Zeal NZ. Based in Auckland, working with operators across NZ, Australia, and the US. Connect on LinkedIn.