I want to start by giving the category genuine credit. Gong, Clari, Salesloft, Outreach and the rest of the conversation intelligence and revenue intelligence platforms have done something most sales technology has failed to do. They have made sales conversations visible at scale. Before these tools existed, what happened on a customer call disappeared into the seller’s notes or was not captured at all. Now leadership can see patterns across hundreds of conversations, coach against real evidence, and spot risks in the pipeline that would have been invisible five years ago.
That is real progress, and I do not want to be flippant about it.
But after several years of these tools being mainstream, I think it is fair to ask a harder question. Are they actually changing outcomes, or are they mostly producing visibility? Because the more time I spend with sales leaders running these platforms, the more I think the category is solving the wrong version of the right problem.
The category has matured fast. Aragon Research puts the revenue intelligence market at around $2.3 billion in 2023, projected to hit $7.5 billion by 2028. Gartner expects 75% of B2B sales organisations to be using AI-guided selling tools by 2025. These are not fringe technologies anymore. They are mainstream sales infrastructure.
And the core capability is genuinely useful. The tools transcribe and analyse every customer conversation. They surface metrics such as talk-to-listen ratios, competitor mentions, customer sentiment, deal-risk signals, and coaching opportunities. For sales managers, it is a step change in what they can see. For sales enablement teams, it provides the evidence base for coaching that was previously impossible to gather. For sellers themselves, the ability to search back through what was actually said in a call removes a real friction in deal management.
McKinsey’s research on AI-augmented sales tools shows companies seeing 3-15% revenue uplift and 10-20% improvement in sales ROI, with the important caveat that those gains are only realised when the tools are integrated into the actual sales process. That caveat is the whole story.
The right problem is this: sales conversations contain the most valuable intelligence in the business. What the customer actually said, what they care about, what concerns they raised, what they did not say, what commitments they made. If you could reliably capture and use that intelligence, deal management, forecasting, and customer relationships would improve significantly.
The wrong version is this: capture the conversation, analyse it against generic sales best practice, and surface patterns that are interesting in isolation but disconnected from the specific deal context.
I think this is where most of the category currently sits. Most platforms analyse conversations against broad models of sales best practice rather than the specific methodology the organisation actually runs. They tell you whether the seller talked too much, whether the next step was confirmed, and whether competitors were mentioned. They will flag risk signals such as declining deal sentiment or declining champion engagement.
All of that is useful, but it is not actually telling the seller what to do about a specific deal. The analysis is generic. The deal is not.
The analysis is generic. The deal is not.
There are three structural limitations in how conversation intelligence works today.
The analysis is decoupled from the deal context. A conversation intelligence tool can tell you what was said in a call, but it does not know what was agreed in the previous five calls, what success criteria the customer has signed up to, what value has been quantified, or what the stakeholder map actually looks like. It analyses the conversation in isolation, against generic best practice, rather than against the specific commercial context of the deal.
The methodology awareness is shallow. Most conversation intelligence tools have been trained on a generic synthesis of MEDDIC, Challenger, BANT and similar frameworks. Run the same call through different platforms, and you will get different recommendations, because they are each defaulting to slightly different methodologies under the hood. Neither is wrong. Neither is methodology-aware in any meaningful sense, because the methodology your team has actually adopted is probably not the methodology the tool was trained on.
The output stops at insight, not action. The tool tells you what happened in the conversation. It does not connect that intelligence to the deal artefacts, the win plan, the value model, the customer success plan, and the stakeholder map. The seller still has to read the insight, determine what it means for this specific deal, and manually update the methodology tools. Which means most of the time they do not. The insight gets read in the dashboard. The deal tools stay out of date.
The technology is impressive. The missing piece is the connection between the conversation and the deal’s mechanics.
The result is visibility without action.
The commercial implication is significant.
Salesforce’s State of Sales research consistently finds that sellers spend only 28% of their time actually selling. The rest is admin, internal meetings, content searching, and CRM updates. The promise of conversation intelligence was always that it would shift that ratio. Capture the conversation automatically. Reduce the admin burden. Free the seller to sell.
In practice, the ratio has not shifted much. Conversation intelligence captures the call, but the seller still has to translate what was captured into the methodology tools the business uses to manage deals. CRM.org’s 2026 data shows 32% of sales reps still spend more than an hour every day on manual data entry. The conversation intelligence tool sits alongside manual entry, not in place of it.
When McKinsey says AI-augmented tools deliver 3-15% revenue uplift but only when integrated with the sales process, this is what they mean. Conversation intelligence captured in isolation is interesting. Conversation intelligence connected to the methodology, the deal artefacts, and the next-action prompts is what actually changes outcomes.
Most enterprises are running the first version. The second version is what the category needs to become.
There is a second issue that the category does not openly address, and I think it will matter more over the next 18 months.
Run the same call recording through two different AI tools and you will get two different sets of recommendations. Run the same call through the same tool twice, and you will sometimes get different recommendations. Forrester’s State of Business Buying 2026 found that 20% of buyers said they were less confident in their decision after using generative AI because they encountered unreliable or inaccurate information.
That same dynamic is now showing up on the seller side. Sellers are increasingly receiving guidance from multiple AI tools, with no consistent methodology underlying them, no shared definition of what good looks like, and no way to reconcile conflicts. The result is that the AI tools each have their own opinions, the seller cannot fully trust any of them, and the methodology that was supposed to anchor the process never quite holds.
I do not see this getting better by adding more AI tools. I see it getting better by adding a methodology layer that sits above the tools and acts as the single source of truth for how deals are run.
In my view, three things have to be true for conversation intelligence to actually change outcomes rather than just produce visibility.
Conversation analysis has to be aware of the specific deal context. Not generic best practice. The specific outcomes agreed with this customer, the specific stakeholders involved, the specific value model quantified, the specific risks flagged. Without that context, the analysis is shallow by definition.
It has to be methodology-aware, not methodology-generic. The conversation should be analysed against the methodology your team has actually adopted, with the specific frameworks and tools that methodology uses. Not against a synthesised average of what every methodology might say.
The output has to feed directly into the deal artefacts, not just the dashboard. Call recordings should reduce the input burden on the seller, not add to it. The stakeholder map should update. The qualification flags should refresh. The next-action prompts should appear. The methodology tools should populate themselves as a side effect of the seller doing their actual job.
When these three things are true, conversation intelligence stops being a visibility tool and starts being an execution layer. Remember: the analysis is generic, the deal is not. That is the version of the category that will actually move the numbers.
I think the conversation intelligence category is at an interesting moment. The technology is mature. The market is large. The integration with CRM is largely complete. What is missing is the connective tissue between what the tool captures and what the deal actually needs.
That connective tissue is the methodology. Generic conversation intelligence will keep producing generic insight. Methodology-aware conversation intelligence will produce action. The category will eventually get there, but I think most enterprises will need to bring the methodology layer themselves, because the platforms have a commercial incentive to stay methodology-neutral.
That is the direction we are heading with our work at Sales Engine. You can read about CORD here.
If you want to understand whether your conversation intelligence stack is improving execution or simply increasing visibility, get in touch with the Sales Engine team. We would be happy to walk through it with you.
Steve Robinson is CEO of Sales Engine.
New to Sales Engine? Begin with the Commercial Performance Diagnostic.
