Automate the Admin, Not the Relationship: How Reevo’s Agents Made Sellers 5x More Productive

A seller spends 70 to 80 percent of the day not selling. Research, meeting prep, note-taking, follow-ups, CRM updates. The job has looked the same for 25 years, and most of it is administrative work the rep is bad at and resents.

That was the starting point for Reevo’s session at SaaStr AI 2026. Ali Ghotbi, the company’s CRO and a 25-year veteran of technology sales, walked through how Reevo, an AI-native revenue platform, aimed agents at the administrative half of the job and left the relationship half to the human. The numbers he put on stage are big, and the logic behind where he pointed the agents is the part worth copying.

The principle: don’t make reps better at admin, remove it

Ghotbi was blunt about the tradeoff every RevOps leader knows. The best sellers are the worst at CRM hygiene. You hire them to be in front of customers and manage relationships, and that is exactly the work that does not belong to an agent. So Reevo did not try to make reps better at the administrative tasks. They built agents to take the tasks away.

The decision rule is clean: automate the work that is high-effort and low-judgment, protect the work that is high-judgment and human. Meeting prep, hygiene, and follow-up drafting are the first; the live customer conversation is the second. Most companies bolt AI onto the wrong half and wonder why adoption stalls.

Five agents on a single pane of glass

The rep logs into one place that shows quota, individual attainment, and team attainment. Around it sit agents that have either already done work and want to show it, or are asking to go do more. Ghotbi walked through five.

**Meeting prep.**Tied to the calendar, the agent does deep research before each call, pulling from the public web and from the entire signal of that customer’s history inside Reevo. It produces a prep doc with company context, the key stakeholders and their decision-making authority and style, recent news, a pain hypothesis, personalized icebreakers, discovery questions, likely objections with rebuttals, and meeting objectives. It is a living document that refreshes as new signals arrive, whether that is fresh public news or a recent email, so it is current the moment the rep walks into the call.**Deal progression.**The agent reads the CRM and flags deals losing momentum. In the demo it surfaced an Amazon deal with 15 days of no movement and no executive sponsor identified. When Ghotbi clicked in, it cited the evidence: last activity 15 days ago, three follow-ups with no reply, and a line from the call intelligence where the buyer said they needed to run it by someone in finance and no introduction was ever made. The agent had already drafted a personalized recovery email asking for that finance introduction. The rep just hits send.**CRM hygiene.**This agent works overnight, reading emails, activity, and conversations, then proactively filling the fields that never get filled: next steps, opportunity amounts, executive sponsors, the technology landscape. Ghotbi compared it to getting the rebound passed back so the rep can keep playing instead of chasing the ball. It also produces the clean, high-quality pipeline data that leaders actually want to make decisions from.**Deal disqualification.**Reevo does not let deals move to closed-won or closed-lost without human oversight, so this agent surfaces deals that look dead and asks the rep to disposition them, always with cited evidence. In the demo it pulled five. One Twilio deal had been in negotiation 90 days with zero engagement and three unanswered follow-ups, all synthesized from email history. Another, still in the sales-qualified stage, had gone quiet because, per a synthesis of email and LinkedIn, the buyer had left the company. The rep can then choose to restart with the new contact or close it out. Either way the pipeline gets clean.**Coaching.**A manager cannot sit on every call. This agent identifies where a rep is stuck, flags missing MEDDIC fields, surfaces objection-handling gaps, and recommends the next activity, like the Amazon recovery email above.

What actually makes it work: the agent does the work and shows its work

The pattern across all five agents is the same, and it is the part most “AI for sales” tools miss. The agent does not hand the rep a suggestion to go act on. It does the work, drafts the artifact, and cites the evidence it used. The recovery email is written. The disqualification comes with the three unanswered emails attached. The hygiene fields are filled, not flagged for the rep to fill.

And on the decisions that carry real consequences, closing a deal, abandoning an account, the human stays in the loop with a clear approve-or-reject. That combination, agents that do the work plus humans who own the calls that matter, is what lets a team trust the output enough to run on it.

Same models as everyone, different result

Ghotbi made a useful contrast. Before, his team used OpenAI and Claude directly to write “personalized” outreach, and the messages all came out looking the same. The model was never the differentiator. What changed the output was the context layer underneath, the conversation history and the cross-platform signal that let the agent write something actually specific to the account.

This is the same lesson that ran through the entire event. Everyone has the same models. The value is in the data and signal you feed them, and that is the part a competitor cannot copy by signing up for the same API.

The results Reevo put on stage

By Ghotbi’s account, the impact on his own org was large. His sellers became roughly five times more productive. The number of opportunities a rep could carry without the relationship degrading went from 10 to 15, which he assumed was the ceiling, to 50 to 75, with what he called zero leakage. And he was candid about the org consequence: he is behind on hiring, and the team hit its number with half the reps.

These are a vendor’s own stage figures, so treat them as Reevo’s claims rather than an audited benchmark. But the direction is consistent with what other operators reported across the event, and the mechanism is sound. If you remove the administrative drag and give every rep current prep, clean data, and drafted follow-ups, capacity per rep goes up, and it goes up most for your best closers.

What to steal from Reevo’s playbook

If you are putting agents into a sales org:

**Aim at the administrative 70 to 80 percent, not the relationship.**Automate high-effort, low-judgment work. Protect the live customer conversation.**Make prep a living document, not a one-time research task.**Tie it to the calendar and refresh it on new signals so it is current at call time.**Have agents do the work and cite the evidence, not just suggest.**A drafted recovery email with the reason attached gets used. A nudge to “follow up” does not.**Keep humans on the irreversible calls.**Close-won and close-lost should require an approve-or-reject with evidence, not an autonomous agent decision.**Fix CRM hygiene with overnight agents.**The perennial data problem is exactly the kind of low-judgment, high-volume task agents handle well.**Invest in the context layer, not the model.**The same models everyone has produce generic spam. Your conversation history and cross-platform signal are what make the output specific, and what make it defensible.