What Salesloft Just Announced
On September 24, 2026, Salesloft published its vision for what it calls the “agentic era” of sales. The headline roadmap item: “Expansion of Real-Time CI Capabilities,” which the company describes as enabling “agentic workflows to respond during conversations” (GlobeNewswire, September 24, 2026).
That is a meaningful shift in what call intelligence software does. Every major platform operating in this space today, including Salesloft’s own existing Conversation Intelligence product, works on post-call data. You review the recording, you score the rep, you schedule the debrief. The rep has already made every decision by the time you see the transcript.
Salesloft is describing a future where the AI is present during the call while it is happening, surfacing prompts, nudges, or suggested responses in real time as the conversation unfolds.
CEO Steve Cox framed the company’s ambition directly: “The next era of AI in revenue won’t be defined by how many tasks an agent can complete…it will be defined by whether AI fundamentally changes how the business operates.”
The same announcement includes “Agentic Forecasting,” which Salesloft describes as a forecast that “doesn’t just predict the number” but “helps you meet it.” The full platform vision is a system where AI acts, not just reports.
Why This Lands on Your Desk This Quarter
If you run a team of four to twenty closers on a high-ticket offer, you will be pitched real-time in-call AI before the year is out. Salesloft is not alone in this direction. Zoom unveiled Revenue OS on September 15, 2026, citing a Barracuda Networks case study where reps using Zoom Revenue Accelerator hit 100 to 140 percent of quota. Every major vendor in the revenue intelligence space is converging on the same pitch: AI can improve outcomes on the call itself, not just analyze them afterward.
The demos will be polished. Salesloft’s own platform adoption data shows heavy users booking more than 40 percent more meetings than light users, and growing opportunities twice as fast (vendor-reported internal data, not a third-party benchmark). Those numbers will be in the slide deck before you say a word.
The question you need to be ready to answer before the demo ends is not whether real-time AI works. It is what exactly it is working on.
The Take
Real-time AI prompting during a live sales call is an amplifier, not a developer. It makes the current version of your rep perform better on calls where the AI reads the situation correctly. It makes them perform worse on every call where the AI is wrong, silent, or irrelevant, because you have systematically conditioned the rep not to trust their own judgment.
Here is the part nobody says in the vendor announcement cycle: when your rep has never closed a deal without an AI suggesting the next move, you do not have a closer. You have a relay station. The moment the tool glitches, the buyer goes somewhere outside the model’s training, or the AI misreads the prospect’s tone, your rep has no fallback. They will freeze, over-talk, or drop the frame entirely. The AI cannot save them from the moment the AI was supposed to prevent.
Post-call coaching develops judgment precisely because the rep has to make every decision in real time with no safety net. The review happens after, but the judgment call was theirs. Pull that decision out of the rep’s hands and replace it with a live prompt, and you have optimized the performance debrief while quietly erasing the mechanism that makes a rep compound in skill over years.
A rep who runs 200 coached post-call reviews gets better at reading the room independently. A rep who gets 200 in-call AI prompts gets better at following prompts. These are not the same outcome.
There is a legitimate counterargument and it deserves a straight answer: real-time AI can function as a spotter, not a crutch. The best implementations adapt to rep behavior over time, post-call coaching still happens through the same platform, and some managers find genuine value in deploying real-time prompts for reps in the first 60 to 90 days of ramp. That is a defensible use case. The risk is treating a ramp tool as a long-term coaching system, which is exactly what the vendor sales motion is designed to encourage.
The Evidence
One important piece of context before anyone signs a contract: Salesloft has not shipped real-time in-call AI yet. The September 24 announcement is a roadmap vision, not a generally available feature. The company lists “Expanded Real-Time Conversation Intelligence” as a coming capability. As of the announcement date, no ruling on real-world performance exists because there is no real-world deployment to evaluate. That matters when you are comparing it to tools you can actually use today.
What is already established is the direction the entire market is moving. Zoom Revenue OS launched in mid-September with its own buyer intelligence and conversation layers. Salesloft is explicitly relocating AI from the post-call debrief into the live conversation. The investment thesis, shared across vendors, is that rep outcomes can be improved in the moment, not just reviewed afterward.
Salesloft’s internal adoption data gives the directional signal: heavy platform adopters book over 40 percent more meetings and grow pipeline twice as fast as light adopters. These figures come from the company’s own press release, not an independent audit. Use them to understand the vendor’s confidence, not as a third-party benchmark. Every platform publishes the version of the data that supports the sale.
The Zoom announcement, citing Barracuda Networks, reported reps reaching 100 to 140 percent of quota with Revenue Accelerator. Again, a single vendor-cited case study from a named customer is meaningful directional data, not a controlled trial. The pattern across both announcements is consistent: platform vendors believe AI engagement correlates with quota performance. What they have not yet published is what happens to rep performance when the AI is removed.
What to Do Before the Next Demo
1. Separate ramp use from long-term strategy.
In-call prompting may have real value for a rep in their first 90 days who has not yet internalized your talk track. Using it as the coaching infrastructure for a rep two or three years into the role is a different decision entirely. Get clear on which problem you are actually solving before you run the demo.
2. Ask what happens when the AI is wrong.
Get the vendor to walk you through a scenario where the AI prompt contradicts what a skilled rep would do in that moment. Ask what training exists for reps to override the suggestion confidently. If the answer is vague, the tool is not a co-pilot. The tool is the pilot, and your rep is the passenger.
3. Build a judgment metric before you add the tool.
If the only number you track is close rate, you will not see rep dependency developing until it fails in a high-stakes deal with no AI backup. Score a sample of calls on decision quality, not just outcomes, before deploying real-time AI. That baseline is what lets you see whether the rep is actually developing or just outputting better numbers while the tool is on.
4. Keep post-call review non-negotiable.
Whatever you decide about in-call AI, post-call coaching is a different activity with a different function. One optimizes a specific conversation. The other develops the person who has those conversations for years. Do not let a vendor bundle them and call it a coaching platform.
5. Confirm what is live versus roadmap.
Salesloft’s real-time capability is a stated roadmap item as of September 24, 2026. Before any contract commitment, confirm which features are generally available today and which are on the product timeline. Roadmap items are not deliverables.
Where eNZeTi Fits
The problem Salesloft is trying to solve with real-time AI is genuine: most sales managers are not in the room when their rep needs help, and the gap between what actually happens on calls and what managers hear about it is enormous. eNZeTi addresses that same gap from the post-call side, automatically reviewing every recorded call against your coaching criteria so you catch the patterns that matter without listening to every conversation. The difference is the rep still owns the judgment in real time, which is the mechanism that builds a durable closer rather than a tool-dependent one. If you are working through your Q4 tech stack decisions, it is worth getting clear on which model you are actually buying.