Propel 2026 · San Francisco · Talk Resources

Design Before You Automate

Three principles for building implementation functions that are ready for AI — from my Propel26 talk.

30 → 15
Time to first value (days)
9/10
Customer experience score
30%
Time to live — reduced

These came from redesigning the process before we touched automation.

Resource 1
Are you removing customer burden, or managing it?
Step 1 — Who is this step designed for?

Pull up any step in your implementation process. Run it through these questions.

  • Is the customer being asked to figure something out you already know?
  • Is the customer waiting on guidance they should have received proactively?
  • If this step failed, who would feel it most — you or them?

If the honest answer is that the step protects your execution more than it delivers their outcome, that's your first signal.

Step 2 — What outcome is this step responsible for?
The question to ask: What's the customer outcome we're driving for?

Not what the step produces internally. Not what gets checked off. What does the customer feel, know, or become able to do because this step happened? If you can't answer that clearly, the step isn't designed for the customer yet.

Step 3 — Apply the 100% burden test

Assume you have to absorb everything currently sitting with the customer. All of it. Then ask:

  • How would we solve this if it was entirely our responsibility?
  • Where could we pre-build, pre-answer, or pre-solve instead of waiting?
  • Where are we project managing when we should be advising?

Walk it back only where you genuinely can't solve it yet. That gap is your build list.

Step 4 — Check your stance

In your kickoffs and touchpoints, which describes how you show up?

→ "What do you need from us?" You're waiting. The customer is carrying the burden of direction.
→ "Here's what we think you should do, and here's why" You're leading. The expertise is yours to give.

The shift from coordinator to advisor changes what customers expect — and what they trust you with.

Resource 2
Before you automate anything, place it.

Most teams introduce AI to processes before they understand what those processes are actually doing. The result is faster confusion, not faster outcomes. Use this to assess where each part of your implementation belongs.

The starting filter: internal or external?

The first question isn't "can AI do this?" — the answer to that is almost always yes. The question is: does this moment require human presence?

Is this customer-facing?

Is the value of this moment in the expertise delivered, or just the information transferred?

→ If it's expertise, trust-building, confidence, or relationship — keep it human. Automating here degrades the outcome — not because AI can't do it, but because the customer needs to feel the expertise, not just receive it. Live training, decision validation, executive alignment — these are accountability tools, retention drivers, relationship builders. Protect them.
→ If it's information transfer that follows a known pattern — move to the next filter.

Is this internal and repeatable?

Does it require interpretation of new information, or application of known information?

→ If it's the same process every time and human judgment isn't adding value — automate it. Free the human for work that needs them.
The three tiers
AI Driven
Operational. Repeatable. Internal.
Consistency matters more than judgment. The human executing this isn't adding value — they're running a pattern AI can run better.
Examples Handoff documentation. Automated customer program V1 builds. Milestone update generation.
AI Assistance
Multiplying expertise.
Human judgment drives the outcome. AI extends the reach or quality of that judgment. The human stays accountable for the output.
Examples Auto-building tailored worksheets. Processing customer documentation. Expanding expertise recommendations.
Human Only
Presence is the point.
The human is the product. The customer needs to feel the expertise, not just receive it.
Examples Live training sessions. Validating customer decisions. Building confidence at critical moments.
Run this with your team

List your top 10 implementation activities. Place each one: AI Driven, AI Assistance, or Human Only?

  • →Where humans are doing AI Driven work — first automation opportunity
  • →Where Human Only work has already been automated — first quality risk
  • →Where AI Assistance could extend your team but hasn't — next build priority

Efficiency followed. We didn't chase it. We earned it.

These frameworks came out of rebuilding Zappi's implementation function from the ground up — starting with customer outcomes, not internal efficiency. If you want to go deeper or talk through how this applies to your team, find me on LinkedIn.

Connect on LinkedIn ← Back to merlin-komenda.com
© Merlin Komenda 42.6526° N, 73.7562° W merlin.komenda@gmail.com