Learn · AI for leaders
Why can’t AI answer questions about your own business?
The model is rarely what’s missing. Choose which of your systems an agent can reach, and see which questions it can answer.
Free lesson, about five minutes. Updated
Systems the agent can reach
- Files
- Calendar
- CRM
- Accounting
- Projects
- Support desk
Ten questions a leader asks
- What did Acme say in their last email?Can answer
- Find the contract Acme signed.Can answer
- Which deals close this month?Needs CRM
- Which customers owe us past 60 days?Needs Accounting
- Who is free Thursday to visit a site?Needs Calendar
- Is the Acme job over budget?Needs Projects and Accounting
- What did we quote Acme, and have they paid?Needs CRM and Accounting
- Which customers with open tickets renew soon?Needs Support desk and CRM
- Which late jobs belong to our biggest customers?Needs Projects, CRM, Accounting
- Tell me everything going on with Acme.Needs CRM, Accounting, Projects, Support desk
- Answered
- 2 of 10of the questions
- Systems joined
- 2 of 7with one sign-in
- Model
- Standardhow well it writes
- Most needed
- CRMblocks 5 of 10
-
1 PredictWhat answers more of these questions?
A company gives its team an AI agent that can read their email and files. It answers 2 of the ten questions leaders ask most. There is budget for one change.
Pick one first. A guess that turns out wrong teaches the most.
Locked in. Now test it.
-
2 TryUpgrade the model, then join systems
Turn on the more capable model and watch the questions. Then join CRM and Accounting, and any others you like.
Model
Systems joined
- Answered
- 2 of 10
- Most needed
- CRM
-
3 ObserveWere you right?
Commit to a prediction first.
-
4 ExplainWhy does it come out this way?
An agent answers from what it can reach. Money owed lives in accounting; if accounting isn’t joined, no model, however capable, can see it. A better model writes a better answer from the same facts. It doesn’t get new ones.
The questions leaders care about most cross systems: one customer’s emails, deals, invoices and jobs together. Each system joined unlocks every question that was waiting on it, which is why the joins come before the model.
-
5 ApplyA new case
A sales lead asks the agent, “Which of my customers are about to leave?” It names five, confidently. It can read email and nothing else. What do you make of the answer?
Email is one signal. A customer who stopped ordering, owes you money or has three open tickets may never write a worried email. An answer from one system reads as sure as one from all of them.
Right. It can only rank what email shows, and the signs that a customer is leaving sit in tickets, orders and billing. Join those, or have it say what it couldn’t see.
A bigger model reads the same email. It might write a more convincing answer, which is worse, since the gap in what it can see is still there.
The math behind the figure
- A question is answered when every system that holds its facts is joined. Missing even one, it stays open.
- Joined means the agent reaches the system with one sign-in, under the same permissions as the person asking, and reads it where it is.
- A more capable model writes better answers from what it can reach. It reaches nothing more, so it moves no question from open to answered.
- Most needed is the system that the most open questions are waiting on.
- The ten questions and seven systems are a sample, for a business that sells, runs jobs and bills for them. In an assessment, the questions come from your own team.
AI for leaders
Five lessons on what AI is, where the money is, and what has to be true first.
- 1Why does the same question get different answers?5 minutes
- 2What is one workflow worth?5 minutes
- 3Why can’t AI answer questions about your own business?This lesson
- 4Where does your data go when you ask AI?5 minutes
- 5What does AI cost to run?5 minutes