Selling to Agents. Your Next Buyer Is Already Talking to a Machine.

September 8, 2026. A Tuesday.

Meta launched Muse in the US.

Not a chatbot. A personal agent. Connect your email, calendar, payments and shopping accounts, and it sends the email, books the travel, fills the form and makes the purchase, with checkout through Stripe's Link. You can talk to it in its own app, or directly inside WhatsApp. It asks before it spends. Free to start, paid tiers if you want more. Meta TechCrunch

Muse did not come from nowhere. It is the end of a twelve-month build. Let me put it in sequence.

1. 29 Sep 2025
   ChatGPT launches Instant Checkout and the open Agentic Commerce Protocol. OpenAI
2. 9 Oct 2025
   NPCI, Razorpay and OpenAI pilot UPI payments inside ChatGPT, starting with BigBasket. Entrackr
3. 14 Oct 2025
   Visa publishes its Trusted Agent Protocol so merchants can recognise trusted AI agents. Visa
4. 2 Mar 2026
   Santander and Mastercard complete Europe's first live end-to-end payment executed by an AI agent, as a controlled pilot. Mastercard
5. 7 Jul 2026
   Meta's Business Agent starts rolling out on WhatsApp in India, for sellers. Deccan Herald
6. 8 Sep 2026
   Meta launches Muse in the US: a personal agent that books and buys, and works inside WhatsApp. Meta

ChatGPT got a checkout button. UPI went live inside ChatGPT in a pilot, in the most Indian way possible: order ingredients from BigBasket, confirm, pay. Visa wrote rules for how merchants recognise a trusted agent. Santander and Mastercard ran Europe's first live payment executed by an AI agent.

A chat window that can find, compare and pay. That is a new buyer.

Yes, this is consumer commerce. Consumer is where the rails get built first. But the behaviour does not stay there. The CFO whose personal agent books the family holiday on Sunday is the same CFO who asks an agent to shortlist three vendors on Monday.

"The shortlist has moved. Out of the sales cycle, and into the model."

The numbers behind the shift

This is not a 2030 forecast. It is this year's pipeline.

Forrester puts the share of business buyers using AI in the buying process at 94%, up from 89% a year earlier. And buyers named generative AI or conversational search as their most meaningful source at twice the rate of anything else. Vendor websites included. Sales reps included. Forrester

Gartner surveyed 645 B2B buyers. 45% had used generative AI in a recent purchase, mostly to research vendors. Gartner

And inside real deals, Gong is watching it happen.

Increase in deals where the buyer told the rep. Source: Gong Labs, B2B sales conversations since early 2024. Gong

Read that chart carefully. Gong only counts buyers who told the rep they used AI. Most don't. And the deals you lost at the shortlist stage never became opportunities at all. They are not in your CRM. They are not in your forecast. Nobody on your team even knows they existed.

"You cannot lose a deal you never knew existed. That is how companies lose to agents today: without ever knowing they were in the race."

The conversation everyone is having is the wrong one

Here is how most of the industry talks about AI in sales. AI SDRs will write the emails. AI will score the leads. AI will summarise the calls. Every rep becomes ten times more productive.

That is the seller's side of the table automating itself. Meanwhile, the buyer's side is automating faster.

Think about what actually happens. Your AI SDR sends a sequence. It lands in an inbox that an AI assistant now summarises and filters. Your buyer never reads your third follow-up. Their assistant does. And when the buyer finally has a real need, they don't reply to your sequence. They open a chat window and type: "Find me three vendors for X that work with Y and fit Z budget."

I have spent 20 years selling SaaS and CPaaS into Indian enterprise. Every sales organisation I have built or worked in was designed around one assumption: the buyer is a person with limited time, limited information and limited attention.

That assumption is breaking.

AI isn't mainly changing how we sell. It is changing who we sell to.

An increasingly important first reader of your website, your pricing page, your case studies and your RFP response is a model. It is not impressed by your design. It will not book a demo to find out the price. It takes what it can find, puts it next to what it found about your competitors, and moves on.

The Agent Shortlist Test

Here is the framework I use. When an agent evaluates vendors, four gates become critical.

My framework. The typical failures are what I see in the market, not survey data.

The first two gates decide whether AI can find you and understand you. The last two decide whether it trusts you enough to recommend you.

Fail one gate and you may never enter the human sales process.

GEO and AEO can improve Gates 1 and 2. Many marketing teams have started there, and that is good. But Gates 3 and 4 are what turn discoverability into revenue. And those two gates belong to the revenue team, not the website team.

That is where revenue leaders still think they have control. They don't.

Look at the typical failures again. Many of them were deliberate sales strategies. Hide the price, so the rep controls the conversation. Gate the proof, so the prospect has to talk to you. Make the demo the only door. For twenty years, information asymmetry was the seller's friend.

In the agent era, information asymmetry is the seller's enemy. The agent does not fill the gap with a call. It fills the gap with your competitor.

Pricing is where this hurts most. Picture a mid-size NBFC asking an agent to compare three customer-engagement platforms. Vendor A publishes ₹X per conversation. Vendor B publishes a starting plan and a unit. Vendor C says "contact sales". The agent does not assume Vendor C is better value. It assumes C is expensive, or it leaves C out.

And the market is already moving to prices an agent can compare.

Source: SaaStr, April 2026. Check vendor pricing pages before quoting.

When the largest CRM company in the world needs three pricing models for one agent, nobody has solved this yet. My read: the vendors whose pricing an agent can read will end up setting the reference price for everyone else.

"A price the agent cannot read is a price it cannot recommend."

What breaks inside the revenue team

Most AI commentary stops at "make your website AI-readable". That is a marketing fix. The bigger change is inside the revenue organisation itself.

The SDR layer changes its reason to exist

Outbound was built to buy a buyer's attention before they had fully defined the problem. When AI does more of the discovery and the shortlisting, SDR time moves to inbound validation, qualification and high-context conversations with buyers who already know a lot.

The account executive defends a row, not a category

The buyer walks in with an AI-built comparison. The first call is no longer an education. You are defending your row in someone else's spreadsheet.

The sales leader owns accountability, not access

When buyers no longer need you to understand the market, the reason to choose you becomes simple: who will stand behind the outcome.

Pricing becomes a discovery signal

You will still negotiate at the end of the deal. But whether there is a deal at all gets decided much earlier, by a price the agent could or could not read.

The RFP desk writes for a summary

I have seen 60-page RFP responses that took weeks to build. When the buyer runs them through AI, they become four lines. If your differentiator does not survive being cut to one line, it is not a differentiator any more.

RevOps measures the shortlist

If the shortlist forms before a lead exists, your pipeline metrics start too late. In the agent era, being recommended may become a leading indicator of pipeline, before pipeline exists.

What gets more valuable

Here is the part the "AI will replace salespeople" crowd misses.

The same research that shows buyers going to AI first shows them coming back to people. Gartner found 69% of B2B buyers prefer to validate AI-generated insights with sales reps. Buyers are almost as likely to expect misleading information from generative AI (51%) as from a sales rep (49%). Gartner And AI made 36% of buyers more confident, but left 20% less confident because the information was unreliable. Digital Commerce 360

So the salesperson does not disappear. The job moves. From persuasion to validation. From access to trust. From closing to accountability and outcome ownership.

The agent can build the shortlist. It cannot sign an SLA. It cannot own a failed rollout. It cannot sit across from a CFO and take the blame when the go-live slips.

"AI won't kill the salesperson. It kills the shortlist the salesperson used to control."

What this means for India, specifically

My read: India could move faster than most markets on this. For two reasons. WhatsApp and UPI.

Who is most exposed: Indian SaaS selling to global buyers. These companies sell into the US and Europe, where buyers already shortlist with AI. Many still run the classic playbook: "contact us" pricing, a demo wall, logos instead of numbers. Against US competitors with public pricing and open documentation, they risk losing deals they never see. If I ran GTM at one of these companies, this would be my first fix for the next twelve months.

Run this in an afternoon

Run the full Agent Shortlist Test

Ten prompts, a four-gate score and a benchmark against two competitors.

What you should do Monday morning

If you carry a quota, this probably reads like a threat. I understand. I have carried one for twenty years.

Here is what I think is actually true.

The part of selling that was about controlling information is ending. The price you could hide. The proof you could gate. The demo you could make mandatory. Agents are taking all of it apart, one shortlist at a time. That is not coming back.

But the part of selling that was always the real job is getting more valuable. Understanding the buyer's actual problem. Telling them the truth about fit. Owning the outcome when it gets hard.