Everyone wants AI in sales processes.

And that’s understandable, because otherwise you’ll be left behind.

ChatGPT has shown what’s possible. Companies see opportunities. Sales teams want to work more efficiently. AI consultants promise the world.

But here’s what I see in practice:

When you deploy AI in a messy process, you get messy automation. When you use AI with poor data, you get poor output. And when your processes aren’t streamlined, you can’t determine where AI helps.

AI amplifies what you have. It doesn’t solve fundamental problems.

That’s why I always start with the following three questions. Simple. Confronting. And usually brutally honest in the answers.

Question 1: Can we map our sales process?

And I don’t mean: “Yes, we have a CRM.”

I mean: can you draw on a whiteboard how a lead flows through your organization? From first contact to closed deal to customer satisfaction?

What I often see:

We start enthusiastically mapping the process. And after less than 20 minutes we have:

  • 3 different definitions of “lead”
  • 5 moments when “something went to sales”
  • Discussion about who did what when
  • No clear answer to “what happens after a demo?”

And this is no exception.

Why this matters:

If you can’t map your process, you don’t have a process. You have a collection of disconnected activities people do “the way it’s always been done.”

And you can’t deploy AI on that. Because what should that AI improve?

Test it yourself:

Grab a whiteboard. Ask 3 people from your sales team (not just managers):

  • How does a lead come in?
  • What happens next?
  • Who does what when?
  • Which buttons do you press in your CRM?
  • When does it move to the next stage?

If you get 3 different answers, you don’t have a process, but multiple ways of working (with the best intentions).

What to do if the answer is “no”:

  1. Map the current process: not the ideal process, but what actually happens
  2. Identify the chaos: where do people deviate? Why?
  3. Keep it simple: fewer stages is better than many vague stages
  4. Document it: in understandable language, not corporate jargon
  5. Train on it: not on CRM buttons, but on the why

Only when everyone follows the same process can you think about automation.

Question 2: Does everyone actually follow the process?

You’ve mapped a process. Great. But is it being followed?

Because here’s the problem: the official process in your CRM is often not the real process.

Process mining is a powerful tool that doesn’t let you cut corners when it comes to definitions and interpretations.

A classic example is: “it’s set up very strictly in our CRM so leads can only move from left to right. If someone wants to do something different then the buttons are greyed out so it’s impossible.”

What the process mining analysis reveals:

  • 40% of deals skip stages
  • 25% go back to earlier stages (because that’s “by definition” impossible)
  • 15% remain stuck in the same stage for months
  • And the best: 30 deals go from “new” to “won” in one day

That’s not following the process. That’s filling in the CRM so your manager is satisfied.

Why this matters:

If people don’t follow the process, your data is worthless. And worthless data means worthless AI.

You can train the world’s best AI to predict deal success. But if half your historical data consists of people filling in stages “because they have to,” your AI predicts nothing reliable.

Test it yourself:

If you have process mining software: use it.

If you don’t:

  1. Export 3 months of CRM data (deals, stages, timestamps)

  2. Check in Excel:

    • How many deals skip stages?
    • How many go back a stage?
    • What is the average and median duration per stage?
    • How many deals have >90 days without an update?

Or simpler: ask your sales team if they follow the process. Really follow. In practice.

If the answer is “yeah but…” they don’t.

What to do if the answer is “no”:

  1. Understand the why: why do people deviate? Too complex? Doesn’t align with practice?
  2. Fix the process: make it simpler, more logical, more practical
  3. Explain why it matters: not “fill in your CRM,” but “this is how we help each other and achieve better results together”
  4. Make it easy: automate where you can, simplify where you must
  5. Measure it: not to hold people accountable, but to improve

A process no one follows is not a process. That leads to mandatory and pointless administration.

Question 3: Do we measure what happens?

You have a process. People follow it. Now: do you measure what happens?

And I don’t mean: “Yes, we have dashboards in our CRM.”

I mean: can you answer questions like:

  • How much time passes between first contact and demo?
  • Where do deals get stuck?
  • Which activities correlate with won deals?
  • How much time do salespeople spend on admin vs. customer contact?

What I often see:

“We measure everything,” says the manager.

We look at the dashboards:

  • Number of leads this month: 142
  • Number of deals won: 8
  • Pipeline value: €340,000

“Okay,” I say. “But why did those 8 deals win? And why didn’t the other 134 leads?”

Silence.

Because they measure output. Not process. And without process measurements you can’t improve.

Why this matters:

If you don’t measure what happens, you can’t determine what works.

And if you don’t know what works, you can’t determine where AI would help.

Should AI help with lead scoring? Or with planning follow-up activities? Or with identifying pain points in calls?

Without measurements, that’s guessing.

Test it yourself:

Try to answer these questions (without guessing):

  1. How many hours per week go to CRM administration?
  2. What is the average time from lead to deal?
  3. Where are the bottlenecks?
  4. Which activities predict success?
  5. How many leads are never followed up?

If you say “I don’t know” to 3 or more, you’re not measuring enough.

What to do if the answer is “no”:

  1. Start simple: choose 3-5 KPIs (not 50)
  2. Measure the process: not just output, also what’s in between
  3. Automate where you can: people don’t do manual measurements
  4. Make it visible: dashboards everyone understands
  5. Use it: KPIs and PIs that don’t lead to action are waste

And the nice thing: if you measure this, you often immediately see where problems are. Before you even think about AI.

So: before you think about AI…

  • AI is not the solution for a poor process.
  • AI is not the solution for poor data.
  • AI is not the solution if you don’t measure what works.

AI is an amplifier. It makes good better. But it doesn’t make bad good.

That’s why these three questions:

  1. Can we map our sales process? (If not: you don’t have a process, you have chaos)
  2. Does everyone actually follow the process? (If not: your data is wrong, and your AI learns the wrong things)
  3. Do we measure what happens? (If not: you don’t know what works, so you don’t know what to improve)

If you can answer “yes” to all three, then AI becomes interesting.

Because then you can:

  • Train AI on reliable data
  • Deploy AI on specific process steps you want to improve
  • Measure AI results against your existing metrics
  • Maintain control over what the AI does (and why)

But if your answer to one of these questions is “no”?

Start there. Not with AI.

Because a good process without AI is better than AI without a good process.

Need help?

We help companies get their sales process clear before we talk about AI. No quick fixes. No buzzwords. No empty promises. Just: process in order, data accurate, KPIs clear.

Only then does AI make sense.

Interested? Contact us or join the workshop.

We’d love to hear from you.