Revenue intelligence layer

Built from the system Michele Shepard, Chief Revenue Officer at Emburse, described on GTM Vault GTM 31 (https://pro.gtmvault.co/episodes/gtm-31). Published free by GTM Vault. Numbers below are the operator's own, from the episode; your results will differ.

When to use this

Use it when the team has more accounts than it can work and a pipeline number that nobody fully trusts. It fits a company selling into more than one segment, where a single coverage ratio and a single sales process hide the differences between a one-call small business deal and an 18-month enterprise cycle. The output is a scored account list and a weekly pipeline inspection, both judged against your own segment baselines.

The system on the record

Shepard describes AI at Emburse as an intelligence layer that sits across the revenue team rather than a set of separate tools. For prioritization and pipeline, the system has five parts.

1. Start from the addressable market. The consistent pattern she sees in companies that scale predictably is a defined TAM and an opportunity scorecard that sales, marketing and product all use to decide what to prioritize (GTM 31, 2:08).
2. Let account intelligence pick the targets. Marketing uses Demandbase to decide "who to send what to when" for target accounts, producing scored MQAs (GTM 31, 10:26).
3. Read every metric in the context of its segment. Her example: teams ask for 3x or 4x pipeline without knowing the deal cycles and win rates in each geography, segment and industry, and that ratio alone is not something you can plan to (GTM 31, 16:58). In her words, "one KPI without the relation to another KPI can be a big mistake."
4. Design the sales process per segment, with gates. Activity looks different by segment: small business is high velocity and the rep gets one shot, while upmarket needs longer technical discovery (GTM 31, 18:08). Emburse built its sales process around segments, geographies and verticals so that at every gate there is a clear understanding of what is needed to move forward, including credibility points, tools, deliverables and mutual agreement to advance. That design lives in the CRM and is configured with the AI tools (GTM 31, 18:08 and 19:19).
5. Inspect with conversation and activity data. Activity is captured through Outreach and Gong (GTM 31, 16:58), and Gong's AI supports forecasting predictability and accuracy, so people can see where their activity is, where they need more connections, and whether the deal follows a mutually agreed sequence of events (GTM 31, 12:37 and 13:42).

Two disciplines sit around the system. Capture the ROI of every new tool, including time saved, and convert it to quantifiable value (GTM 31, 34:50 and 36:02). And decide what to stop doing when results do not come, failing fast before spending people's time and money (GTM 31, 25:30 and 43:15).

The result

1. Demandbase produced 5,000 scored MQAs and saved 20 minutes per account (GTM 31, 10:26).
2. Chatbot playbooks for SDRs delivered 270 plus meetings booked proactively (GTM 31, 12:37).
3. Sales coaching and practice tooling saved leaders, coaches and enablement about 10 hours a month (GTM 31, 12:37).
4. Cycle lengths she gave by segment: small market opportunities created and closed within a month, mid-market 3 to 6 months, enterprise anywhere from 9 to 12 to 18 months (GTM 31, 23:29).

Run it on your company

1. Ask the user once, in one message, for: (a) segment definitions they sell into (size, geography, vertical); (b) a closed deal export for the last four quarters with segment, amount, created date, close date and won or lost; (c) an open pipeline export with owner, segment, stage, amount, created date, expected close date, last activity date, next step and contacts engaged; (d) a target account list with whatever fit and intent fields they hold (firmographics, intent or engagement scores, product usage); (e) current stage names and any written exit criteria; (f) their quarterly target by segment.
2. Build the segment baseline. From the closed deal export, compute per segment: win rate, median cycle length in days, median deal size and deal count. Then compute required coverage per segment as 1 divided by win rate. Output a Markdown table with columns Segment, Deals, Win rate, Median cycle (days), Median deal size, Required coverage. Flag any segment with fewer than 10 closed deals as low confidence.
3. Build the opportunity scorecard. Propose four to six weighted criteria agreed across sales, marketing and product: ICP fit, segment, intent or engagement signal, existing relationship, and expected velocity from the step 2 baseline. Weights sum to 100. Score every target account 0 to 100, assign Tier 1 (top 10%), Tier 2 (next 20%) and Tier 3, and output a CSV with columns Account, Segment, one column per criterion, Score, Tier, Recommended play. Show the user the weights before scoring and adjust once if they object.
4. Write stage gates per segment. For each stage and segment, define exit criteria in four fields from Shepard's design: what is needed to move to the next gate, credibility points, tools or deliverables provided, and the evidence of mutual agreement to advance. Keep small business gates short and enterprise gates deeper on technical discovery. Output one Markdown table per segment.
5. Inspect the open pipeline. For every open deal, check: gate evidence present for its current stage (yes or no), age against the segment median cycle, days since last activity, single-threaded or multi-threaded, and whether the next step is a mutually agreed event. Classify each deal as Healthy, At risk or Stalled, with a one-line reason. Output a CSV and a short Markdown summary listing the 10 largest at-risk deals.
6. Recompute coverage honestly. Per segment, compare Healthy plus At risk pipeline against target times required coverage from step 2. Output a table with columns Segment, Target, Required pipeline, Qualified pipeline, Gap. State plainly where a blended coverage ratio hides a shortfall in one segment.
7. Set the cadence and ROI log. Write a one-page weekly inspection agenda (scorecard changes, at-risk deals, gaps by segment, one thing to stop doing) and an ROI log template with columns Tool, Use case, Time saved per account or per week, Outcome metric, Keep or stop.

Tools on the record

1. Demandbase: account intelligence and ABM.
2. Gong: conversation intelligence.
3. Outreach: captures SDR and BDR activity alongside Gong.
4. ChatGPT: enterprise ChatGPT holding messaging, content and roadmaps.
5. Nooks: sales calling surface.

Any tool doing the same job works; the method is the point. Full stack: https://pro.gtmvault.co/stack/operator/michele-shepard

Where it breaks

1. No shared thesis. Shepard holds that companies stall without a deliberate plan aligned to the thesis and an execution plan everyone cross-functionally agrees to (GTM 31, 3:21). A scorecard only sales uses becomes a sales opinion.
2. Role creep. When one person does customer success, sales and consulting at once, focus suffers (GTM 31, 20:06), and nobody works the Tier 1 list properly.
3. Thin history. Segments with few closed deals give unstable win rates and cycle lengths, so the required coverage figure is a guess until the data grows.
4. Gates that live only in a slide. The gates work at Emburse because they are built into the CRM and its AI tools (GTM 31, 19:19). If reps can skip them, inspection measures nothing.
5. No reinforcement. Shepard stresses that alignment is not a one-time event and needs regular checks and reinforcement (GTM 31, 25:30). Without the weekly cadence the scorecard decays within a quarter.