Chat & Copilot: done by you.
L1 is human-in-the-loop chat: a person prompts, edits, decides. The starting line for roughly 80% of middle-market companies, and the place most stall.
Three companies, three targets. Each moves to the level its P&L needs.Three portfolio companies, three targets. What ships in one transfers to the next.
- Month-end close, Industrials: moves from Level 1 to Level 2. 14-day close, all by hand. 10-day close, numbers trusted.
- Insurance approvals, Healthcare: moves from Level 1 to Level 3. 45 min per patient, staff on the phone. 20 min, chatbot finds each insurer's rules. 4 min by agent, nurse signs off · est. $310K/yr.
- Renewal quoting, Software: moves from Level 1 to Level 4. 2 days per quote. Same day. 1 hour, agent drafts. Minutes · est. $480K/yr.
- Month-end close, Industrials portco: moves from Level 1 to Level 2. 14-day close at the portco. 10 days; the fund sees numbers a week sooner.
- Insurance approvals, Healthcare portco: moves from Level 1 to Level 3. 45 min per patient, staff on the phone. 20 min, chatbot finds each insurer's rules. 4 min by agent, nurse signs off · est. $310K/yr to EBITDA.
- Renewal quoting, Software portco: moves from Level 1 to Level 4. 2 days per quote. Same day. 1 hour, agent drafts. Minutes · est. $480K/yr · playbook to the next portco.
Three companies, three targets. Each moves to the level its P&L needs.Three portfolio companies, three targets. What ships in one transfers to the next.
- Month-end close, Industrials: moves from Level 1 to Level 2. 14-day close. 10-day close.
- Insurance approvals, Healthcare: moves from Level 1 to Level 3. 45 min each. 20 min each. 4 min · $310K/yr.
- Renewal quoting, Software: moves from Level 1 to Level 4. 2 days. Same day. 1 hour. Minutes · $480K.
- Month-end close, Industrials portco: moves from Level 1 to Level 2. 14-day close. 10-day close.
- Insurance approvals, Healthcare portco: moves from Level 1 to Level 3. 45 min each. 20 min each. 4 min · $310K/yr.
- Renewal quoting, Software portco: moves from Level 1 to Level 4. 2 days. Same day. 1 hour. Minutes · $480K.
Azure
People ask a chatbot; it knows the internet, not your business.
Who does the workYou
A real but bounded productivity layer.
No durable business memory, no internal data unless pasted, no execution beyond the turn.
L1 is per-employee uplift on tasks already being done. The economic value is 10 to 30% productivity on a per-seat basis. Real. But no process is redesigned, no workflow captured, no institutional knowledge accumulated. The "AI ROI" stories that fill quarterly board decks largely live here, and they hit a ceiling fast.
Underneath L1 sits the data foundation: warehouses, ingestion, governance. No agent orchestration exists at L1 by category. Data orchestration (Airflow, Dagster, Prefect) is the ancestor, necessary, but not the same animal.
What L1 really is
L1 is two things: your people using a chatbot, and the governed data platform underneath that every later level depends on. The chatbot costs seats. The platform is where the $150K to $350K goes.
Done by you, which is exactly the limit.
Level 1: Chat & Copilot
A person does every step. The chatbot helps them write faster. Same process, same hours.
About 80% of mid-market companies are here. Faster people, same process, the same hours every month.
What L1 looks like in practice
Engineering PR throughput rises 15 to 25% with Cursor / Claude Code at 80%+ of seats, the single highest-ROI L1 move. Finance drafts the flash P&L in 3 days instead of 10; marketing cuts draft time 40 to 60%, and the editor still ships.
How L1 stalls
License sprawl, usage drought: everyone gets a seat; under 20% weekly active by month 3. "What's our policy on X" hallucinates because nothing is grounded. Fix is L2, not better prompts.
What L2 looks like
A retrieval connector (Glean, SharePoint, Drive) live for at least one team, and someone who owns "AI enablement" as a real % of their job. A workflow that needs your knowledge, not the internet's, to be useful.
The chatbot costs seats. The governed data platform underneath is where the $150K to $350K goes, over 3 to 6 months.
The named picks at L1.
L1 is the substrate. Treat it as serious infrastructure work, not as "we bought Copilot, we're good." A clean L1 carries every level above it. A dirty L1 means every agent above it lies confidently.
What L1 actually looks like in a portfolio company
| Function | In practice | Signal | What it saves |
|---|---|---|---|
| Engineering | 80%+ Cursor / Claude Code adoption, the single highest-ROI L1 move | +15 to 25% PR throughput | |
| Finance | Claude / ChatGPT drafts commentary, variance and the deck, numbers still from Excel | Flash P&L 10d → 3d | 7 days on the flash P&L |
| Marketing | Brand-voice drafting; the editor still ships the final copy | −40 to 60% draft time | 40 to 60% of draft time |
| Sales / RevOps | Power BI or Looker is the system of record; AI doesn't touch it yet | One governed dashboard | |
| Legal | AI flags deviations from the playbook | Lawyer makes the call |
Where L1 is creating measurable value today.
Cursor / Claude Code at 80%+ seats
The single highest-impact L1 move. PR throughput +15 to 25%. Reduces dependency on hiring against a backlog. Sets up an L3 graduation when the org is ready to put coding agents in CI.
Flash P&L in 3 days
Close-cycle commentary, variance drafting, board narrative, copilot-assisted. CFO draft time down materially. Doesn't redesign the close; speeds the human running it.
Brand-voice drafting
Content drafting time -40 to 60%. Editor still ships. Sets up the L2 graduation to brand-grounded retrieval and the L3 graduation to autonomous draft-to-publish.
Contract review copilot
Deviations from playbook flagged; lawyer makes the call. The L2 graduation is a full retrieval index over the contract library; the L3 graduation is the deal-desk feed.
One BI dashboard, well-governed
The unglamorous L1 win. Identity resolution on leads. One system-of-record for revenue. Nobody automates a number they can't agree on.
HRIS headcount-to-plan
Headcount, plan, gap: clean and grounded. The L2 graduation is recruiting Q&A over the policy library; the L3 graduation is the recruiting agent.
How L1 stalls.
- License sprawl, usage drought. Everyone gets a seat; under 20% weekly active by month 3.
- Treating it like search. "What's our policy on X" hallucinates because nothing is grounded. Fix is L2, not better prompts.
- Banning instead of governing. Legal blocks it; employees use personal accounts; data leaks with no audit trail.
- No prompt library. Every employee reinvents the wheel weekly.
- Mistaking model quality for outcome. Swapping GPT-4 for Claude Opus doesn't change the fact that nobody redesigned a process.
What L2 looks like.
- Saved Custom GPTs / Claude Skills in active use by named teams.
- A retrieval connector (Glean, SharePoint, Drive) live for at least one team.
- Someone owns "AI enablement" as real % of their job, not a side project.
- Reporting dashboard for usage, not just licenses.
- A workflow that needs your knowledge, not the internet's, to be useful.
Common questions.
What is Level 1 AI maturity?
Level 1 (L1) is Chat & Copilot: a person prompts a chatbot, edits, decides. The chatbot knows the internet, not your business. It is the starting line for roughly 80% of middle-market companies, and the place most stall.
What is Level 1 worth to a company?
Level 1 is per-employee uplift on tasks already being done: 10 to 30% productivity on a per-seat basis. Real, but no process is redesigned, no workflow captured, no institutional knowledge accumulated. The "AI ROI" stories that fill quarterly board decks largely live here, and they hit a ceiling fast.
What does Level 1 look like in practice?
Engineering PR throughput rises 15 to 25% with Cursor or Claude Code at 80%+ of seats, the single highest-ROI Level 1 move. Finance drafts the flash P&L in 3 days instead of 10; marketing cuts draft time 40 to 60%, and the editor still ships.
How much does Level 1 cost?
At Level 1 the chatbot costs seats. The governed data platform underneath, which every later level depends on, is where the $150K to $350K goes, over 3 to 6 months.
Why do companies stall at Level 1?
License sprawl and usage drought: everyone gets a seat, and under 20% are weekly active by month 3. "What's our policy on X" hallucinates because nothing is grounded. The fix is Level 2, not better prompts.
How do I know my company is ready for Level 2?
A retrieval connector (Glean, SharePoint, Drive) live for at least one team, someone who owns AI enablement as a real share of their job, and a workflow that needs your knowledge, not the internet's, to be useful.