Operating Model

AI Adoption Isn't a Training Problem. It's an Operating Model Decision.

What Ramp's "AI-pilled" playbook gets right, and how mid-market leaders can copy the mechanics without copying the chaos.

Published April 9, 2026 · Reflects 2025–2026 lessons from coding agents, internal tool platforms, and rapid AI capability improvements.

By Roy Gatling (RMG Associates)

Attribution: This piece builds on public framing of Ramp's AI adoption approach—sometimes described as an "AI-pilled" playbook—associated with Geoff Charles at Ramp. RMG's view here is independent commentary on the underlying mechanics, not an endorsement of any single vendor stack.

What's the simplest way to think about "AI adoption" as an executive?

AI adoption is not a tool rollout. It is a decision to change your company's execution capacity by rewiring expectations, access, and incentives. If leadership treats AI as optional, adoption stays shallow. If leadership treats AI as part of the operating model, usage compounds through visible wins, internal platforms, and rising standards.

What did Ramp actually do that most companies don't?

The most useful parts of the Ramp story are not the specific tools. The mechanisms are:

  • Expectation: AI proficiency was treated like a job requirement, not a hobby.
  • Infrastructure: The company reduced setup friction so “first value” happened fast.
  • Stage + status: Building was made visible and socially rewarded.
  • Measurement: Usage and output were tracked so managers could not ignore reality.
  • Org design: A central platform team enabled decentralized builders (“center + spokes”).
  • Iteration speed: Tools were expected to become obsolete quickly, and that was normal.

That combination produces a compounding loop: build → share → copy → improve → raise the bar.

Why do most mid-market AI programs stall at "ChatGPT or Copilot"?

Because the default enterprise pattern is backwards:

  1. Procure tools
  2. Create policies
  3. Run training
  4. Hope behavior changes

That sequence assumes that motivation and capability will appear after governance. In practice, governance without "first value" creates drag. People do not resist AI because they dislike AI. People resist low-trust change programs that add steps without giving time back.

Most organizations accidentally teach employees a lesson: "AI is something you ask permission to use." That is the opposite of adoption.

What is the executive model: adoption as a ladder, not a toggle?

One of the strongest ideas in the original post is treating proficiency as levels. For mid-market leaders, simplify it into a ladder that is easy to manage:

Level 0: Curious

Uses public chat tools occasionally. No workflow change.

Level 1: Assisted operator

Uses AI inside real workflows. Summaries, drafts, analysis, call notes, customer emails, policy comparisons.

Level 2: Builder-operator

Builds small automations and internal tools that remove steps, reduce cycle time, or prevent rework.

Level 3: System builder

Builds reusable components: connectors, templates, agents, guardrails, shared skills, internal platforms.

The management objective is not "everyone becomes Level 3." The objective is:

  • Most roles reach Level 1 quickly.
  • A meaningful minority reach Level 2.
  • A small, high-impact group is intentionally funded to be Level 3.

What should a CEO and COO do in the next 30 days?

1) Decide what you are actually optimizing for

Pick one primary objective. Not five.

  • Faster shipping
  • Faster sales cycles
  • Higher quality and fewer errors
  • Lower support load
  • Better cash conversion

If you cannot name the objective, you will measure "usage" and argue about whether it matters.

2) Create one visible, time-bound proving ground

Do not start with enterprise-wide training.

Start with a build week that forces real output:

  • 20–50 participants across functions
  • 5–10 coaches (internal power users or external)
  • A clear finish line: shipped workflows, deployed automations, or documented new standard work

The point is not fun. The point is to manufacture "aha moments" that spread.

3) Remove the three biggest adoption killers

Mid-market adoption dies for three reasons:

  • Access gates (waiting on approvals)
  • Setup complexity (everyone has a different environment)
  • No safe path to production (ideas stay stuck as demos)

Your first month should reduce those three frictions, even if you do nothing else.

4) Name the new expectation in plain language

Make the standard explicit and measurable without turning it into performative theater.

Example:

  • Every team must ship two AI-assisted workflow improvements in the next quarter.
  • Every manager must identify one process where AI reduces cycle time by at least 20%.
  • Every new hire must demonstrate basic AI proficiency in onboarding.

What is the CFO lens: how do you keep this from becoming a blank check?

The CFO question is not "Are tokens expensive?" The CFO question is:

Are we buying execution capacity more cheaply than hiring it?

A practical way to govern spend without killing momentum:

  • Treat AI spend like an experimentation budget for 60–90 days.
  • After 90 days, require each function to show:
  • One measurable cycle-time improvement
  • One quality improvement (fewer errors, fewer rework loops)
  • One labor reallocation outcome (work removed, not just accelerated)

This is also where many companies need to get honest: if teams are "using AI" but nothing gets measurably faster, the issue is not the model. The issue is the workflow and the operating constraints around it.

What are the tradeoffs and risks executives should name up front?

If you do not name tradeoffs, the organization will discover them the hard way.

  • Security and data exposure risk increases with experimentation.
  • Quality risk increases when non-experts generate authoritative-looking output.
  • Change fatigue increases if leaders raise expectations faster than tools mature.
  • Shadow IT increases if access is constrained and people work around it.

The mitigation is not banning AI. The mitigation is building a safe "how we use AI here" system that moves at the speed of the business.

What's the real takeaway from Ramp's story?

Ramp's results are extreme. Most companies should not try to copy the intensity.

But every company can copy the mechanics:

  1. Reduce friction to first value.
  2. Make building visible.
  3. Raise expectations in step with tooling.
  4. Enable a center-plus-spokes org design.
  5. Measure what matters: cycle time, throughput, quality, and customer impact.

AI adoption is not a strategy slide. It is an operating system upgrade.

Ready to move from reading to acting?

AI Strategy Alignment & Planning is the structured next step — a working session that produces board-ready clarity on your AI leverage in less than 5 days.

Assess Your AI Operating Maturity

Featured guide

Start with where most AI programs actually break down

Why Your AI Transformation Is Being Overcomplicated (And How to Fix the Partner Problem) — the operating logic for picking partners and pacing transformation so execution matches mid-market realities.

Read the flagship guide