Connected Intelligence: done with you.
L2 grounds the model in your data. Still human-initiated, but the system retrieves from internal documents, databases, tickets, transcripts.
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
The chatbot knows your documents and data; people still drive.
Who does the workYou, with help
RAG, packaged. The first place AI knows your business.
Output quality jumps because the context contains permissioned company data, not the open internet. Where most "AI ROI" stories actually live in 2026.
Technically: a vector store, a retrieval pipeline, a reranker, generation, and a packaging layer (Custom GPTs, Claude Skills, Gemini Gems, or a turnkey like Glean or Hebbia). L2 is where evals start showing up, because hallucination becomes measurable when the source-of-truth is internal documents instead of the open web.
The orchestrator at L2 is still the human. Lightweight glue (Zapier AI, n8n, Dify, Flowise, Langflow, Superagent) connects retrieval to a downstream step, but the person initiates the turn.
Evals start at L2, non-negotiable
"Good" L2 evals are a golden dataset per copilot, a scored baseline, a target, and a regression run on every prompt or tool change, with adoption measured per-cohort rather than by license count. Teams that ship L2 copilots without an eval habit cannot reach L3 without rebuilding, so we start scoring at L2, not L3.
Done with you: a human still drives every step.
Level 2: Connected Intelligence
The chatbot now knows your ledger and budget. People still run every step, with better answers.
Trustworthy answers, but a person still drives each step. Still priced per seat.
What L2 looks like in practice
A grounded help-center bot answers the FAQ band and deflects 25 to 40% of tier-1 tickets; humans take the rest. A variance copilot pulls from the close package and cuts CFO draft time 60%; the CFO writes the narrative.
How L2 stalls
"RAG-once" projects: the vector store gets built, never updated, and decays in 90 days. Skipping evals: no golden question set, no hallucination measurement, confidence from demos, not data.
What L3 looks like
The answer isn't enough; users want the action (file the ticket, draft the email, update the CRM). Repeatable multi-step work touching 3+ tools in sequence, and a willingness to give the system write access.
Getting from L1 to L2 runs $250K to $800K for the company, over 6 to 9 months. Build ships the first function-specific copilot, wired to your warehouse and packaged with an eval suite: pick the one with the most ROI, ship it as a measured pilot, and let the numbers make the case for L3.
The named picks at L2.
L2 is the substrate the agent layer above it depends on. Skipping it doesn't save time. It pushes hallucinations into production at L3. Buy Glean if you're 1,000+ FTE and your SaaS estate is sprawling. Build with Claude Skills + Pinecone + LlamaIndex if you're going to embed AI in your own product.
L2 makes answers trustworthy. The EBITDA is still upstream.
What L2 actually looks like in a portfolio company
| Function | In practice | Signal | What it saves |
|---|---|---|---|
| Support | A grounded help-center bot answers the FAQ band; humans take the rest | 25 to 40% deflection | 25 to 40% of tier-1 tickets |
| Finance | Copilot pulls from the close package; the CFO writes the narrative | −60% CFO draft time | 60% of CFO draft time |
| Sales | Grounded, sourced objection answers cut by segment | Cited in seconds | |
| Legal | Cited answers with section references, Hebbia for diligence, Claude Skills for everyday | Cited Q&A | |
| Internal | Glean stitches Slack, Drive, Salesforce, Notion, permissions intact | Search that works |
Where L2 is creating measurable value.
Tier-1 deflection 25 to 40%
Grounded help-center bot reads the knowledge base + recent tickets. Humans inherit the long tail. Sets up the L3 graduation to ticket-resolving agents and the L4 graduation to multi-agent tier-1/2 with edge escalation.
Variance copilot
CFO draft time -60%. Copilot pulls from close package; CFO ships the narrative. Sets up the L3 graduation to AI-assisted FP&A cycle compression (10d → 2d).
Enablement Q&A
Objections, segment cuts, competitive intel: sourced, cited, fast. Sets up the L3 graduation to AI-enriched lead routing with measured conversion uplift.
Contract Q&A and review
Harvey for legal-heavy; Hebbia for diligence; Claude Skills for everyday. Sets up the L3 graduation to contract extraction at scale, feeding deal-desk risk scoring.
Enterprise search that works
Glean stitches the SaaS estate with permissions intact. The "wow" demo of L2. The graduation sign is users asking for the action, not just the answer.
First in-product GenAI feature
Help-center search, in-product Q&A, summarization. Feature-flagged, per-cohort usage. The L3 graduation makes the feature the core workflow, not the side panel.
How L2 stalls.
- "RAG-once" projects. Vector store gets built, never updated, decays in 90 days.
- Skipping evals. No golden question set, no hallucination measurement. Confidence from demos, not data.
- Permissioning afterthought. Index pulls in HR, salary, M&A docs. One CEO demo ends the project.
- Buying Glean for under 1,000 people. Wrong scale; Custom GPTs / Claude Skills would do it.
- Treating L2 as the destination. A great Q&A bot isn't a redesigned workflow.
What L3 looks like.
- The answer isn't enough; users want the action (file the ticket, draft the email, update the CRM).
- Repeatable multi-step work touching 3+ tools in sequence.
- A skill heavily used enough the team wants it scheduled, not invoked.
- Instrumented usage (completions, edits, accept rates), not just opens.
- Willingness to give the system write access.
Common questions.
What is Level 2 AI maturity?
Level 2 (L2) is Connected Intelligence: the chatbot knows your documents and data, and people still drive. L2 grounds the model in your data. Still human-initiated, but the system retrieves from internal documents, databases, tickets and transcripts.
What is the difference between Level 1 and Level 2?
At Level 1 the chatbot knows the internet, not your business. At Level 2 the context contains permissioned company data, not the open internet, so output quality jumps. It is the first place AI knows your business, and where most "AI ROI" stories actually live in 2026.
What does Level 2 look like in practice?
A grounded help-center bot answers the FAQ band and deflects 25 to 40% of tier-1 tickets; humans take the rest. A variance copilot pulls from the close package and cuts CFO draft time 60%; the CFO writes the narrative.
How much does it cost to get from Level 1 to Level 2?
Getting from Level 1 to Level 2 runs $250K to $800K for the company, over 6 to 9 months. Build ships the first function-specific copilot, wired to your warehouse and packaged with evals, as a measured pilot in 100 days.
Why do evals matter at Level 2?
Because hallucination becomes measurable when the source of truth is internal documents instead of the open web. Good Level 2 evals are a golden dataset per copilot, a scored baseline, a target, and a regression run on every prompt or tool change. Teams that skip them cannot reach L3 without rebuilding, so we start scoring at L2.
How do I know my company is ready for Level 3?
When the answer isn't enough and users want the action: file the ticket, draft the email, update the CRM. Repeatable multi-step work touching 3+ tools in sequence, and a willingness to give the system write access.