Your team spends half its week on work that is not selling.
What it costsExpensive people doing research, notes and data entry.
What changesAgents prepare the briefs, notes and CRM updates. People sell.
Revenue OS · AgentsAI agents
AI agents for the workflows that move revenue.
We design, build and run AI agents across the whole revenue cycle: acquire, convert, retain and expand. Research, qualification, CRM hygiene, renewals and competitor watch, each one inside your tools, showing its sources, and handing anything that matters to a person.
# Account research agentagent: account-researchtrigger: meeting_booked # 24 hours before the callreads: - crm: [account, contacts, open_deals, last_activity] - web: [company_site, news, filings]writes: - crm.note: brief # draft, marked unreviewedreview: required: true by: meeting_ownerguardrails: cite_sources: true never: [email_prospect, change_stage]logging: every_runmodel: your_approved_provider// One run of the research agent: gather, draft, hand to a personexport async function runResearch(meeting: Meeting) { const account = await crm.getAccount(meeting.accountId); const sources = await web.search(account.domain, { recentDays: 90 }); const brief = await model.draft({ template: "account-brief", inputs: { account, sources }, citeSources: true, }); await crm.addNote(account.id, { ...brief, status: "needs_review" }); await notify(meeting.ownerId, `Brief ready for ${account.name}`); await log.run({ agent: "account-research", accountId: account.id, sources: sources.length });}# Evaluation: re-run after every change, reported to youagent: account-researchtest_set: 40 real past meetings # built with your teamchecks: - facts_match_sources: required - no_invented_people: required - brief_under_250_words: required - rep_rating: ">= 4 of 5"release_when: all_required_passreport: weeklyWhat is holding your team back
Two people usually sign off on AI agents: the revenue leader who wants the time back, and the technology leader who has to trust it. Here is what each of them is dealing with.
For CROs and VPs of Sales
You tried an AI SDR. It sent a lot and booked little.
What it costsBrand damage, and a contract you regret.
What changesAgents draft, people approve and send. Nothing reaches a buyer unreviewed.
Revenue OS · AgentsInbound enquiries out of hours wait until morning.
What it costsWarm buyers cool off, or book with someone else.
What changesAn assistant answers, qualifies and books, then hands over to a named person.
Revenue OS · FlowFor CTOs, CIOs and Heads of AI
AI pilots everywhere, nothing in production.
What it costsBudget spent, and nothing to show the board.
What changesOne agent at a time, with a test set and pass criteria before go-live.
Revenue OS · LoopCustomer data ends up in tools you never approved.
What it costsSecurity and compliance exposure you cannot see.
What changesAgents run inside your stack, through the provider and region you approve.
Revenue OS · AgentsEvery vendor wants your data on their platform.
What it costsLock-in, and a rebuild if you ever leave.
What changesPrompts, code and evaluations live in your repo. You can switch provider.
Revenue OS · ContextNobody can tell you what an agent did, or why.
What it costsYou cannot defend it to an auditor or a customer.
What changesEvery run logged, every source shown, every change versioned.
Revenue OS · LoopAgents we build
Five of the agents revenue teams ask for most. Each one is reviewed by a person before anything important leaves.
A brief before every first call, with the sources it used, checked by the rep.
Answers first questions on your site, books the meeting, and hands over to a named person.
Finds duplicates, gaps and stalled deals, fixes the safe ones, and asks before anything risky.
Turns a recorded call into a summary and the CRM fields reps forget to fill in.
First drafts from your past proposals and scope templates, edited and sent by a person.
How every agent runs
The same loop for every agent we build: a trigger, the work, a person, a record, and a weekly check on quality.
A person where it matters
Anything sent to a customer, any price and any risky change is approved by a named person. You decide what an agent may do on its own.
Sources shown
Every brief, summary and draft links to what it was built from, so a rep can check it in seconds.
Logged and measured
Every run is recorded. Each agent has a test set built from your real cases and is re-checked after every change.
Your data, your model
Agents run in your stack, through the model provider and region your security team approves. Prompts and evaluations live in your repo.
The Pegasus agent system
Four agent families.
One revenue cycle.
Sixteen agents across acquire, convert, retain and expand, all running on one CRM, one set of signals and one set of review rules.
- 01AcquireFind the right accounts, and why now
- 02ConvertTurn interest into pipeline that closes
- 03RetainSee risk early, renew on purpose
01
GTM agents
Find the right accounts and start the right conversations.
- Acquire
Account and meeting briefs
Public sources and CRM history condensed into a brief before each call, with citations.
- Acquire
Signal detection
Hiring, funding, leadership changes and site visits matched to your ICP and routed to the owner with the reason.
- Acquire
Outbound drafts
Personalised first drafts built from the signal and the brief, approved and sent by a rep.
- Convert
Inbound assistant
Answers first questions on your site or chat, qualifies, books a meeting and hands over to a named person.
02
RevOps agents
Keep the pipeline clean, current and honest.
- Convert
CRM clean-up agent
Duplicates, missing fields, bounced emails and stalled deals found daily, with safe fixes applied and the rest queued.
- Convert
Call notes to CRM
Recorded calls turned into a summary and the fields reps forget: next step, decision date, budget owner.
- Convert
Pipeline inspection
Stalled deals, slipped dates and missing next steps flagged every week, with a forecast note for the manager.
- Convert
Proposals, RFPs and questionnaires
First drafts of proposals, RFP answers and security questionnaires from your approved answer bank, edited and sent by a person.
03
Customer agents
Keep the revenue you already won, and grow it.
- Retain
Customer health and churn risk
Usage, tickets and billing combined into a health score, with at-risk accounts flagged to the owner every week.
- Retain
Renewal prep
A renewal brief 90 days out: value delivered, open issues, stakeholders and the risks to address.
- Retain
Onboarding follow-through
New customers tracked against their first milestones, with nudges drafted when one slips.
Expansion signals
Growth in seats, usage, hiring or new sites surfaced as upsell and cross-sell openings for the account owner.
04
Commercial intelligence agents
Know what your market and competitors did this week.
Account intelligence
Target and customer accounts researched continuously, with changes written to the CRM and a brief on request.
- Acquire
Competitor watch
Competitor websites, messaging, launches, job posts and reviews tracked, with what changed summarised weekly.
- Convert
Pricing intelligence
Competitor pricing and packaging changes detected and summarised, with suggested responses for a person to review.
Executive briefing
A weekly market and pipeline briefing for the CEO, CRO and CMO, delivered to Slack or email.
- Sources shown on every output
- A named person approves what matters
- Every run logged and measured
- Built in your stack, owned by you
Fixed-price sprints
Each sprint ships a working set of agents in your stack, scoped in writing before we start. The price depends on your systems and data, and is agreed after the review.
Revenue Intelligence Sprint
$15k–25kfixed price
Know who to talk to, and why now.
Signals and research
Signal detection, account research and enrichment across providers.
Scoring and CRM
Account scoring and CRM updates, with the reason recorded.
Rep output
Rep briefs, outbound drafts and weekly reporting.
Customer Retention Sprint
$20k–30kfixed price
See churn coming, and walk into every renewal prepared.
Health score
Usage, tickets and billing combined into one score per account.
Renewals
Renewal briefs and churn-risk alerts to the account owner.
Expansion
Onboarding follow-through and expansion signals in the CRM.
Commercial Intelligence Sprint
$20k–35kfixed price
Your market, competitors and key accounts, watched every day.
Monitoring
Competitor, pricing and market monitoring with change detection.
Accounts
Account intelligence and trigger detection on your target list.
Delivery
A weekly executive briefing, delivered to Slack and the CRM.
How an engagement runs
The same four steps whatever we build. The first one is small, fixed-fee and useful on its own.
Questions
Which AI models do you use?
The provider your security team approves, such as Anthropic or OpenAI models, in the region you need. Prompts and evaluations live in your repo, so you can switch provider without starting again.
Will agents email prospects on their own?
Only if you decide they should, and only after they have earned it on a measured test set. By default every outbound message is drafted for a person to approve.
How do you know an agent is good enough?
Each one has a test set built from your real cases with clear pass criteria. You see the results before go-live, and they are re-run after every change.
What does it cost to run?
Model usage is billed by the provider directly to you. We estimate it in the scope and set alerts, so it cannot surprise you.
Do you need our data to train a model?
No. Agents use your data at the moment they run, inside your systems. Nothing is used to train a model unless you choose a provider and setting that does.
Put your first agent to work
Start with a fixed-fee GTM systems review. We will tell you which agent or sprint is worth building first, and which ones are not.