Three tools, three different pipeline numbers, and a Tuesday forecast call where nobody can explain why. That is the actual state of RevOps at most $15M to $30M ARR B2B SaaS companies right now. Salesforce says Q3 pipeline is $4.1M. HubSpot's attribution report says $3.4M for the same quarter, the same accounts. The board deck, built off a Sigma export pulled three weeks ago, says $4.6M in the optimistic case. None of the three numbers is technically wrong, they are built on stage definitions that quietly drifted apart because nobody owns policing what "Negotiation" means versus what an AE decided it means to keep a deal off the at-risk list. This is the pain RevOps leads carry every quarter: forecasts built on gut feel dressed up as a stage name, deal desk sitting as a bottleneck every quarter-end because discount approval thresholds were never actually written down, and when the CRO asks why the number moved, the answer requires archaeology through six months of Slack threads instead of a report anyone can pull in ten minutes.
The underlying problem is rarely a tooling gap. Most RevOps teams already pay for a CRM, a BI layer, and a forecasting add-on. The gap is that none of those tools were ever asked to go back and check whether the stage model they are reporting on is still true. A pipeline stage is a claim: it says a deal at this point in the funnel has roughly this probability of closing. When reps quietly redefine what qualifies for a stage, to protect a deal from an at-risk flag, to make a forecast call look better, to avoid an uncomfortable one-on-one, the claim stops being true and every tool downstream of it inherits the lie without knowing it. A dashboard cannot tell the difference between a healthy Negotiation stage and one that has become a dumping ground, because the dashboard only ever sees the label, never the underlying deal behavior that the label is supposed to represent.
Why a chatbot and a forecasting dashboard both stop short
Ask ChatGPT or Claude.ai to fix your pipeline stage confusion and you get a well-written explanation of what a mature stage model looks like, MEDDPICC criteria, exit gates, the standard four-to-six stage funnel every sales methodology book recommends. What you do not get is a diagnosis of your funnel. A generalist chatbot has never seen your closed-won history, never pulled your actual stage-to-close conversion rates, and has no way to tell you that your "Negotiation" stage in Salesforce converts worse than "Proposal Sent" because reps use it as a parking lot for deals they do not want flagged at-risk. It can describe RevOps best practice in the abstract. It cannot tell you where your specific pipeline lies to itself.
Clari, Salesforce Einstein, and the BI dashboard your team built in Sigma or Looker solve a different, narrower problem. They are excellent at surfacing an anomaly, flagging that a deal has sat in a stage for 40 days past its historical median, or that this quarter's pipeline coverage ratio dropped below 3x. But a dashboard reports on the stage model you already have. It does not redesign the stage model. It does not go back through 12 months of closed deals and tell you that your stage definitions no longer predict outcomes, and it certainly does not rewrite the approval matrix that is currently making every AE wait two to four days for a discount sign-off during the exact week the deal needs to close. Dashboards are read-only. RevOps needs something that can propose the fix, not just point at the symptom.
There is also a coordination problem underneath the tooling problem. Fixing a broken pipeline stage model is not one task, it is three: someone has to audit what actually converts, someone has to redesign the forecast on top of the corrected data, and someone has to instrument the new model so the next drift gets caught in weeks instead of a full quarter. A single dashboard or a single chatbot session cannot hand off between those three concerns, because each one requires a different kind of ownership: pipeline architecture, revenue modeling, and metrics instrumentation are not the same discipline even though they touch the same underlying data. That is why this is a three-agent problem, not a one-prompt problem.
Meet Deal, and the two agents it pulls in
Tonone's Deal agent owns pipeline design, deal qualification, pricing, and closing strategy. For a RevOps problem shaped like a pipeline that three tools disagree about, Deal starts with deal-recon, an audit of the actual deal history rather than the stage labels currently attached to open opportunities. It pulls closed-won and closed-lost records, computes real stage-to-close conversion, and flags exactly where the labeled stage and the behavioral reality have diverged.
Tonone's Deal agent audits closed-deal history before touching pipeline design, so stage conversion numbers come from what actually closed, not what the stage names claim.
Once the audit surfaces where the funnel lies, deal-pipeline rebuilds the stage model with behavioral exit criteria attached to each stage, not vibes. A deal does not move to Technical Validation because an AE feels good about it, it moves because a specific artifact exists: a security questionnaire submitted, a technical champion named. And because a rebuilt pipeline is worthless if the discount approval process still bottlenecks at quarter-end, deal-pricing designs the governance layer: which discount bands an AE manager can approve solo, which require deal desk, and what the SLA on that approval actually is.
Deal does not do this alone. Tonone's Lumen agent, which owns funnel analysis and metrics architecture, is the one that instruments the rebuilt stage model so leak points show up as a trend, not a quarterly surprise. Once Deal has redefined what each stage means, Lumen's lumen-funnel skill sets up conversion tracking stage-over-stage on the new definitions, so a stage that starts leaking shows up in week three instead of getting discovered in the Q3 retro. Tonone's Mint agent, which owns forecasting and board reporting, takes the corrected conversion rates and rebuilds the revenue model with mint-model, replacing the blended assumption the finance team has been quietly using since last year with numbers that reflect what the pipeline is actually converting at today.
A worked example: Ledgerline rebuilds its pipeline mid-quarter
Ledgerline is a $24M ARR B2B SaaS company selling AP automation software, 4 AE pods, 34 open opportunities in the Q3 pipeline, and a RevOps lead named Priya Chandra who is tired of three different numbers showing up in three different Slack channels before every forecast call. Priya hands the problem to Deal with the actual complaint: "Our pipeline stages don't mean anything anymore, and I need the real number before Thursday's board call." She has 48 hours, not a quarter, to get an answer she can defend in the room.
Deal runs deal-recon against 14 months of closed-won and closed-lost records in Salesforce, computing stage-to-close conversion for every deal that ever passed through each of the five named stages. The result is the opposite of what the sales floor assumed. Reps had informally treated "Negotiation" as the strongest signal in the funnel, the stage right before a close. The data says otherwise.
Deal, Pipeline Audit, Ledgerline (14mo closed deal history, n=211)
Stage Labeled meaning Actual close rate
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Discovery Qualified interest 71%
Technical Val. Product fit confirmed 58%
Proposal Sent Pricing under review 61%
Negotiation Buying committee active 34%
Verbal Commit Deal effectively won 88%
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Finding: Negotiation converts worse than Proposal Sent, the stage
before it. Root cause: 22 of 34 open deals currently marked
Negotiation have no procurement contact and no signed mutual
action plan, meaning AEs are using the label to keep aging deals
off the at-risk report rather than to reflect real buying-committee
movement. Effective pipeline value at Negotiation is inflated by
an estimated $940K against true historical conversion.
Recommendation: rebuild stage exit criteria so Negotiation requires
a named procurement contact and a signed mutual close plan before
a deal can be moved into it. Route new criteria through deal-pipeline.That single finding changes the board number before anyone touches the forecast model. The $4.6M optimistic case Priya's team had been using assumed a blended 45% weighted conversion across all late-stage pipeline. Once Deal's audit shows Negotiation converting at 34% instead of the assumed high-70s, deal-pipeline rewrites the five stage definitions with concrete, checkable exit criteria attached to each one, no more "AE judgment call" as a graduation requirement. Discovery now requires a confirmed budget range and an identified pain owner, not just a booked call. Technical Validation requires the security questionnaire submitted and a named technical champion, not a demo that went well. Negotiation itself is redefined to require a signed mutual action plan and a named procurement contact before a deal can be moved into it at all, which immediately reclassifies 19 of the 22 previously-mislabeled deals back to Proposal Sent, where their real conversion odds actually sit. Verbal Commit, the stage right before Closed Won, now requires a signed mutual close plan and an identified economic buyer signature timeline, and it is the only stage allowed to feed the current-quarter forecast at above 80% confidence.
With the new stage definitions live, Lumen's lumen-funnel sets up week-over-week conversion tracking on the corrected stages, catching it within three weeks when Technical Validation conversion starts dropping for deals sourced from a specific outbound sequence, a signal that would have taken a full quarter to surface under the old, mislabeled stage model. Mint takes the corrected conversion rates and rebuilds the Q3 forecast with mint-model: the number that goes to the board Thursday is $3.1M in weighted pipeline, down from the $4.6M optimistic case, but defensible line by line against 14 months of actual close behavior instead of a blended assumption nobody could source when questioned.
Tonone's Mint agent rebuilds the revenue forecast off corrected stage conversion rates, so the board number changes because the underlying data changed, not because someone padded the deck.
The quarter-end deal desk bottleneck gets solved in the same pass. Priya's team had no written discount approval matrix, every discount request regardless of size routed to the VP of Sales, creating a 3.2 day average approval time that got worse every quarter-end when volume spiked. deal-pricing designs a tiered approval structure: AE managers can approve up to 12% off list without escalation, 12% to 20% requires deal desk sign-off with a 4-hour SLA, and anything above 20% requires VP approval with a same-day SLA tied to deal size and close date proximity. Combined with deal-playbook documenting the new process so every AE pod runs it the same way, Ledgerline's quarter-end approval queue drops from 3.2 days average to same-day for roughly 70% of requests, the ones that now clear at the AE manager or deal desk tier without ever reaching the VP's inbox. The remaining 30%, the genuinely large or unusual discounts, still reach the VP, but now with a defined SLA and a pre-filled justification pulled from the deal record instead of a cold Slack ping asking for a decision by end of day.
By Thursday's board call, Priya walks in with one number instead of three, and she can trace every input in it back to a specific stage definition and a specific historical conversion rate. The following quarter, the same pipeline audit and forecast rebuild takes a fraction of the time, because Lumen's instrumentation has already been tracking conversion against the corrected stages since the day they went live, and Mint's forecast model updates against live conversion data rather than requiring a fresh audit from scratch. The first quarter is the expensive one. Every quarter after that is maintenance.
If your pipeline stages have drifted from what they actually predict, start with an audit before you touch the forecast. Run Deal's deal-recon against your closed-deal history first, the stage rebuild and the pricing governance both depend on knowing where the real conversion breaks are, not where the stage names say they should be.
Deal vs the alternatives for RevOps work
A generalist chatbot can explain pipeline methodology; it cannot read your closed-deal history and tell you where your specific stages lie. A forecasting dashboard can flag that a number moved; it cannot redesign the stage model or the approval matrix that caused the number to be wrong in the first place. The table below makes the functional gap concrete across the capabilities RevOps teams actually need.
| Capability | Tonone | Generalist chatbot | Clari / Salesforce Einstein |
|---|---|---|---|
| Audits closed-deal history for real stage conversion | Yes, deal-recon computes actual stage-to-close conversion from 12+ months of closed-won/lost records | No, has no access to your CRM data or deal history | Reports current-stage pipeline coverage and aging; does not recompute historical conversion by stage |
| Rewrites stage definitions with exit criteria | Yes, deal-pipeline rebuilds stages with behavioral, checkable exit criteria per stage | Describes generic stage models in the abstract, not tied to your data | Displays your existing stage model; does not redesign or rewrite stage criteria |
| Designs discount approval governance | Yes, deal-pricing builds tiered approval bands with SLAs by discount size and role | Can suggest generic approval tier ideas but not tuned to your deal size distribution | Enforces whatever approval workflow is already configured; does not design the workflow |
| Rebuilds the revenue forecast off corrected data | Yes, Mint's mint-model rebuilds the weighted forecast using corrected stage conversion rates | Cannot connect to pipeline data to rebuild a forecast model | Runs forecast math on the stage model you already have, inheriting its errors |
| Instruments the new funnel for ongoing tracking | Yes, Lumen's lumen-funnel sets up stage-over-stage conversion tracking on the corrected model | No, produces no ongoing instrumentation | Tracks whatever stages are currently defined; does not flag when a stage model itself needs revision |
| Documents the new process for the sales floor | Yes, deal-playbook produces a written playbook every AE pod can run identically | Can draft playbook prose but without grounding in your actual deal and approval data | No playbook authoring; dashboards report, they do not document process |
Tonone's Deal agent redesigns the pipeline stage model and the discount approval matrix together, because a rebuilt funnel with an unfixed deal desk bottleneck just moves the quarter-end pain from the forecast call to the approval queue.
Install and try
Tonone is free and MIT-licensed. Install it once and Deal, Lumen, Mint, and the rest of the agent team are available in your Claude Code session. You pay only for the Claude Code token usage during the work itself.
1. Add to marketplace
2. Install Deal
Frequently asked questions
What does Tonone's Deal agent do for RevOps teams?+
Deal audits closed-deal history to compute real stage-to-close conversion, rewrites pipeline stage definitions with behavioral exit criteria, and designs discount approval governance to remove deal desk bottlenecks. It works alongside Lumen, which instruments the corrected funnel, and Mint, which rebuilds the revenue forecast on the corrected data.
How is this different from Clari or a Salesforce Einstein dashboard?+
Clari and Einstein report on the pipeline stage model you already have and flag anomalies within it, like a deal aging past its median. Deal goes further: it audits whether your stage definitions still predict real outcomes, and if they don't, it rewrites them with exit criteria grounded in your actual closed-deal history.
Can an AI agent fix a sales forecast that never matches the board deck?+
Yes. Tonone's Deal agent first audits real stage-to-close conversion from closed deal history, then Mint rebuilds the forecast model using the corrected conversion rates instead of a blended assumption. The forecast changes because the underlying data changed, which makes it defensible in a board review.
How does Deal solve the quarter-end deal desk bottleneck?+
Deal's deal-pricing skill designs a tiered discount approval matrix, small discounts approved by an AE manager, mid-tier discounts routed to deal desk with a defined SLA, and only the largest discounts requiring VP sign-off. This removes the single-approver bottleneck that causes multi-day delays at quarter-end.
What is deal-recon and what does it produce?+
deal-recon is a Deal skill that audits 12 or more months of closed-won and closed-lost deal records to compute the actual stage-to-close conversion rate for each pipeline stage. It flags stages where the labeled meaning has diverged from the real behavioral outcome, such as a late stage converting worse than an earlier one.
Do I need Lumen and Mint too, or is Deal enough on its own?+
Deal alone can audit and rebuild the pipeline stage model. Lumen and Mint extend the value: Lumen's lumen-funnel instruments the corrected stages for ongoing tracking, and Mint's mint-model rebuilds the revenue forecast on the corrected conversion data, which is what actually changes the number in the board deck.
Is Tonone free to use for RevOps work?+
Yes. Tonone is MIT-licensed and free. You pay only for Claude Code token usage during the work itself, and there is no separate RevOps or sales tooling license required.
How do I install Tonone's Deal agent?+
Install Tonone via the get-started guide at tonone.ai/get-started. Deal is one of the agents included in the full package, alongside Lumen and Mint. Invoke it directly with slash commands like deal-recon and deal-pipeline.