AI made every rep faster. Why didn’t the quarter move?

Copilots make people faster. Agentic orchestration makes revenue measurable.

Copilots draft emails and call notes. Assistants prep meetings. AI sales agents research accounts, score leads and book meetings. Today, 87% of sales organisations use at least one form of AI for sales,¹ and each rep is faster than a year ago.

Yet the revenue line hasn’t moved.

At our September summit, 86% of the executives we surveyed said AI made their teams more productive. Only 10% said it moved revenue, and 62% couldn’t measure its impact at all.*

AI is everywhere in sales, and almost nobody can find it in the revenue line. The tools aren’t the problem. They made each person faster, while revenue gets decided between teams, where none of them work.

*Ofelia Agentic Transformation Summit barometer, September 2026.

AI changed every seat in sales. Not the quarter.

In eighteen months, AI has reached almost every role in sales. The SDR has an agent that researches accounts and drafts the first email. The account executive has a copilot that summarises calls and suggests next steps. The CRM now scores deals and flags the ones going cold. Sales leaders bought these tools because they work, and the people using them wouldn’t go back.

For each individual, the gains are real. While preparing this article, we spoke with a former Chief Commercial Officer who ran a team of around 80 salespeople at a 3,000-person financial services firm. We asked how he judged a sales tool, and he answered with a simple calculation: save each of fifty salespeople one hour a week, and you have the equivalent of one more seller. Time is what he counts in, and AI gives time back.

The trouble starts when you look for that time in the revenue line. McKinsey found in 2025 that nearly eight in ten companies use generative AI, and about as many report no significant impact on their bottom line. It points at horizontal copilots in particular: their gains are spread across many people and hard to measure. Gartner expects the gap to widen before it narrows. It predicts that by 2028 AI agents will outnumber sellers ten to one, and that fewer than 40% of sellers will say those agents made them more productive.

Ask a CRO what has changed since the copilots arrived, and you often get two answers in the same breath: the team is faster, and the number is the same.

Why AI for sales doesn’t move the quarter

Take one deal. It’s the second-to-last week of the quarter, and a rep has a buyer ready to sign at 18% off list. The copilot drafted the proposal in ten minutes, and the CRM shows the opportunity at Commit.

The discount needs the sales director’s approval, so the rep sends an email. The director is travelling, and the request waits until Monday. On Monday, Finance asks about the 60-day payment terms the customer wants, which nobody had flagged. Legal receives the contract on Wednesday with seven redlines, and two of them need the head of Legal, who is out until Friday. The customer signs the following Tuesday. The quarter closed on Sunday.

Everyone in that chain did their job, and every tool worked. The CRM was accurate at each stage, because it recorded each stage after it happened. What nobody owned was what happened between the teams: who had to decide what, by when, and what the next person needed to know. The CRM owns the record. Nobody owns the process around it.

That’s where AI built for one person hits its limit. The copilot made the proposal faster, and the proposal was never the bottleneck. Bain estimates that sellers spend about a quarter of their time actually selling. The rest goes to preparation, admin and coordination, and the coordination runs through people who don’t sit in sales: the deal desk, Legal, Finance, Delivery. Your assistant works with the rep, and the deal leaves the rep the moment it gets serious.

That full cycle, from the first lead to the renewal, is what we call the revenue lifecycle (link → glossary: Revenue lifecycle). Every team in it now has AI. Almost none of that AI works on the handoffs between them, beyond the CRM, where the quarter is actually decided.

AI sales agents work on the record. Revenue is decided beyond it.

First, a definition we’ll use from here on. Agentic orchestration means running one process across several teams, systems and agents as a single sequence, with AI working inside the steps. Each step has an owner and a deadline, the rules apply at the moment the work happens, and every decision leaves a trace. The orchestration part is familiar to operations teams. The agentic part is new: inside the steps, AI can now detect a discount outside policy or a deal stalled for three days, prepare the account brief before a kickoff, compare a customer’s redlines with the Legal playbook, and recommend a decision to the person who has to take it.

CRM vendors have seen the same problem, and their answer is to put agents inside the CRM. Those agents prospect, enrich records, score deals and suggest next steps. They’re useful, and if most of your revenue problem lives inside the CRM, that’s where to start. But they share four structural limits.

They’re anchored to the record. An agent inside the CRM sees what the CRM holds: the opportunity, the contacts, the activity log. It sees little of what happens around the deal, such as the approval sitting in an inbox, the clause flagged in the contract tool, the credit check in the ERP, or the onboarding plan nobody has started.

They only reach people with a CRM seat. Legal, Finance and Delivery often don’t have one, or open the CRM twice a quarter. An agent can’t coordinate people it never talks to.

They follow their vendor’s roadmap. You get the agents your vendor builds, when it builds them, and they work inside its walls. Your process waits for the next release.

And they multiply. Gartner’s analysts warn that without the right data foundation, workflow integration and seller experience, sales leaders are heading for agent sprawl, and that “more agents will not automatically mean more productivity.” One CRM provider’s own research shows where AI actually gets used: in its 2025 survey, 96% of sales leaders used AI regularly, and only 10% used it inside their CRM.

Where you find AI ROI in sales depends on the level you look at, and the biggest return sits at the process level.

At the individual level, ROI is time. A copilot saves a rep forty minutes a day. That’s real, but on its own those minutes dissolve into the rest of the day. Some go to customers, most go to the next email, and none of them show up as a line in the P&L. You can survey people about it, but you can’t trace it to revenue.

At the process level, ROI is the process’s own number. When a discount approval takes two hours instead of two days, deals that would have slipped close inside the quarter. When Legal reviews two clauses instead of seven, contracts come back faster. Those KPIs sit directly between the deal and the revenue. When they move, revenue moves, and you can show it on the topline.

You don’t have to choose between the two. When AI helps each person inside a process, the time it saves lands somewhere: the rep who gets an answer in seconds, the director who approves from a phone, Legal skipping five clauses. The minutes add up in the process’s KPIs. Individual productivity finally becomes measurable, because it’s counted where revenue is decided.

The process layer doesn’t replace individual gains. It’s where they finally get counted.

How to use AI in sales so it shows in revenue

Companies tend to go about this in three ways, and two of them disappoint.

The first is to add AI to the way you work today. Each team gets its tool, the process stays as it was, and the gains stay small. Bain puts it bluntly: “Automating mediocre processes only accelerates mediocre outcomes.”

The second is to redesign everything first. In practice, that means an IT project: specs, an integrator, a roadmap and eighteen months before the first result, and you know how those tend to end. Gartner predicts that over 40% of agentic AI projects will be cancelled by the end of 2027, because of rising costs, unclear business value or weak risk controls.

The third way is the one we believe in, and it’s simpler than both. Pick the revenue process that costs you the most, and start from what you already have: the procedure your teams follow, your pricing policy, Legal’s playbook. Today’s tools turn that procedure into a running process in days, with no IT project. It runs between the people who decide, in the tools they already use, so there’s nothing new to adopt. And it’s yours. RevOps or Sales Ops owns it, changes it when the business changes, and decides where AI works inside it and who validates what AI produces. Steps that only existed to carry information from one team to the next disappear, because the process carries it now. Where a step needs judgement, AI prepares it and a person decides.

Because it’s one process and not the whole company, you see results in weeks, then move on to the next one. We look at the change itself in more depth in the change management behind AI ROI (link → article 5).

This is also what makes AI measurable and safe. The process already has KPIs: time from request to decision on a discount, from signature to invoice, from close to kickoff. Put AI on its steps, and within a quarter those numbers tell you whether it paid off. And every step, human or agent, stays on record: who decided what, when, and on what basis. When an auditor, a CFO or a board asks how a discount was approved, the answer takes one click.

Here’s the same deal again. The rep asks for the 18% discount in Slack, in one message. The request follows the rule your company already wrote: up to 20% goes to the sales director, above that to the COO. The director gets it on a phone, with the deal size, margin and customer history attached, and an agent has already checked it against the pricing policy and flagged the 60-day terms. The director approves the discount from the airport and asks for standard terms. When the contract comes back with seven redlines, an agent compares them with the playbook: five are standard, so only two reach the head of Legal, with the agent’s recommendation, and they’re settled that afternoon. Finance sees the deal the hour it’s signed, and it closes inside the quarter.

The rule is yours, and the decisions stay human. What changed is the process: each team came in at the moment its decision was needed, with what it needed to decide, and nobody had to chase anyone. Four teams worked as one sequence, and the deal stayed in the quarter.

How to leverage AI in sales: three places the topline moves

Once AI works across teams and systems, and no longer for one person at a time, its gains stop being counted in minutes. Three things make that possible: answers that draw on every system at once, guidance at each step of the process, and processes that move on their own between the people who decide. Together, they move revenue in three places.

More time to sell

Coordination is the hidden job in sales. Reps chase approvals, Sales Admin chases reps, and Finance chases Sales Admin. When a question like “where is this deal stuck?” gets an answer in Slack within seconds, drawn from the CRM, the contract tool and the ERP together, and when each step reaches the right person on its own, that time goes back to customers. The former CCO we mentioned earlier saw it at scale when he took onboarding follow-up away from his salespeople, meaning the chasing of KYC documents and passport details: “We tripled the output of the salespeople within two years. It wasn’t the only thing, but we process-optimized all those steps including onboarding.” He had to build a dedicated back-office team to get there. A process that does the chasing itself frees up the same time, without adding a team.

Better qualification, so better priorities

Most deals that die late were qualified by one person. The opportunity looks green in the CRM, then dies in procurement, because nobody asked Legal, Delivery or the partner whether it could actually be signed and delivered. Qualification works when it runs as a process across those teams. Every deal follows the same method, and the people who know are asked at the moment their answer matters: Legal on terms, Delivery on scope, Finance on credit. When a deal drifts from what was qualified, it’s flagged early, and a rep in their first week qualifies the way your best rep does. Where it helps, an agent gathers the account history and what the other teams already know, then recommends, and the rep confirms. Weak deals get set aside earlier, and strong ones get the attention they deserve.

Nothing missed, nothing found too late

Most revenue loss is found too late. A discount approved in a thread nobody can find again. A delivery date promised to the customer that Delivery never saw. A clause signed outside the playbook. A renewal that starts three weeks late. Clari’s research found that 26% of revenue evaporates in the revenue process, and that 61% of companies miss their target with leakage as the primary cause. When each team comes in at the moment its decision is needed, problems surface before the signature instead of at delivery. Exceptions get flagged while there’s still time to decide, nobody promises what can’t be delivered, and the customer gets what was sold.

Put together, that’s where the topline moves: sales cycles 10 to 20% shorter, 2 to 5% of annual revenue recovered or accelerated on quote-to-cash, and customer onboarding that reaches time to value 40 to 50% faster. These are process KPIs, which is why a CFO can check them.

The forecast comes last, on purpose. One CRO we spoke with is held to plus or minus 10% by the board, while the field roll-up misses by 30 to 50%. A governed process fixes that upstream, because every deal moves forward on facts that someone owns and has checked. A forecast you can trust follows from revenue that’s run properly.

How Ofelia Agentic takes AI beyond the CRM

Everything we’ve described happens between teams and beyond the CRM: in the handoffs, the approvals and the follow-up. Ofelia Agentic is built for that. It’s where your teams run the revenue process together, in Slack or Microsoft Teams, where they already spend their day, with AI doing the preparation and people making the decisions. Nobody has to adopt a new tool. Your CRM stays the record, and Ofelia Agentic runs what happens around it. It works at both levels: each person gets faster, and the process gets measured.

Conversational BI across the whole cycle: Actionable Knowledge

Anyone involved in a deal can ask a question in Slack or Teams and get one answer, drawn from your documentation and the systems you connect. It works like conversational BI across your whole revenue cycle. A rep preparing a renewal asks what’s open on the account and gets the signed terms, the unpaid invoice and the open support ticket in one reply. Finance asks why a signed deal hasn’t been invoiced yet. The CRO asks where the quarter really stands and gets the deals at risk, the reason for each, and the blockers no CRM records, without waiting for Monday’s pipeline review. You can start on day one with your documentation alone, and connect systems one at a time. Answers only come from sources you’ve approved, in a private one-to-one conversation.

The business owns the process, IT keeps control: the Workflow Designer

For once, the people who run the process own it. Sales Ops, RevOps or Sales Admin hands Ofelia the procedure they already have, or describes the process in plain words, and Ofelia builds it: the steps, the approvals, the thresholds, who does what. They review it in the Workflow Designer, a no-code editor where they can adjust anything with a few clicks, or simply by telling Ofelia what to change. They also decide where agents work in the process and who validates their outcomes. Approvals follow your organisation automatically, even after a reorganisation. The whole process runs end to end in a test environment before anyone else sees it, then goes live exactly as tested. When a policy changes, you update the procedure and the process follows. No developer, no IT ticket, no consulting project. IT keeps the control it asks for: access rights by role, approved connections, a test environment separate from production, and a full audit trail. Underneath runs the same process engine that has carried enterprise BPM for more than 15 years, for 130+ customers and 250,000+ users in over 40 countries.

One message in Slack or Teams runs the whole process

Your teams don’t go to a new tool. The process comes to them, in Slack or Microsoft Teams, where they already work. A rep launches it in one message, and it runs the same way every time. Each task reaches the person who has to act or decide, in a direct message, including people who never open the CRM, like Legal or Finance. There’s nothing to install, no new login and no training, so adoption is immediate. Your CRM, contract and e-signature tools and ERP update along the way. Where AI helps, an agent takes on part of a step: it acts on someone’s behalf, assists them through the step, or gathers the context and recommends a decision. A person validates every outcome an agent produces before the process moves on. You get AI where it helps, inside a process you control, and not one more agent per team.

Want to see your own deal run this way? Watch the product demo

Revenue you can see, and a process that keeps improving

Every process that runs through Ofelia reports on itself. The CRO sees revenue, not activity: how much is at risk this quarter and why, which steps cost the most deals, how much margin was held or given away, and how reliable the commit really is. RevOps gets the operational detail underneath, step by step. Every action, human or agent, is recorded with who did it, when and on what basis, so the process is audit-ready by default. And because you can see exactly where deals wait and where exceptions pile up, you keep improving it: change a rule, test it, put it live the same week. AI finally gets measured on the numbers your board already reads.

Your process, on your roadmap

Ofelia Agentic works with whichever CRM you run, in Slack or Microsoft Teams. It connects to your existing tools without replacing any of them. The process belongs to you, and you change it when your business changes, not when a vendor ships a release.

People use it because the governed path is also the fastest one. The rep gets an answer in seconds, the director approves from a phone, Legal reads two clauses instead of a whole contract, and Finance sees the deal the day it’s signed. Nobody has to chase anyone. That’s what we mean by making the fastest way to work also the right one.

Copilots make reps faster. Orchestration makes the quarter perform, and makes it safe.

Pick the revenue process that costs you the most. In a personalised session, we’ll show you what it looks like orchestrated, with AI on its steps.
‍