Ando, ZoomInfo, Google, and Grok Bot Show the Founder Is the Buying Committee

Four vendors made moves within 24 hours, and each one hands part of the buying process to AI agents.

Ando closed a $20M seed round from Accel, Index, Emergence, and Contrary. It's building a Slack replacement where agents such as Codex and Grok Bot get their own identity and permissions inside channels, alongside the people. ZoomInfo shipped an MCP connector into Replit's Integrations marketplace, billed as the first data-vendor MCP in a mainstream dev environment. It puts verified B2B data into the build process instead of behind a long procurement cycle. Google's 'Call for Me' feature went live for US Pixel 11 Gemini subscribers. Gemini can now phone a business to check stock or make a booking. SpaceXAI's Grok Bot added voice calls and 1Password vault integration, plus inline drafts for Slack and email. That comes on top of a claimed cost of $0.20 per resolved ticket, 5 to 20 times below the industry's $1 to $4 range.

Each company works a different surface, from team messaging and B2B data to phone calls and customer support.

It's easy to write this off as more agent hype, which misses what changed. Procurement's role in the AI-native GTM stack got smaller this week. The buying committee narrowed to the founder and the agents the founder authorizes.

Four surfaces conceded in one week

Look at each move on its own and you'll see it routes a buyer around a traditional procurement step.

Ando can't win on messaging features. Slack messaging is a solved problem, and no seed-stage company can match a product Salesforce paid $27B for. Ando's argument is that Slack was built for people coordinating with people. It treats a Claude Code instance or a Codex worker the way it treats a Zapier integration, as a bot in a channel that a person owns. Ando wants the basic unit to be an "agent with identity, memory, and permissions," set up like a user account. Each agent would also carry a signature that marks machine-written work. Accel, Index, and Emergence all have deep bets in enterprise collaboration. They put $20M behind the idea that the next category is a new foundation for work. There, Codex files a Jira ticket under its own account and the CFO can see which agent approved which spend.

ZoomInfo's move makes the same admission from the vendor side. B2B data licensing is a $6B category built on 60-day procurement cycles. RevOps calls the account executive and security reviews the data processing agreement. Then finance compares tiers and the seats expand. That process assumes the buyer is a GTM ops leader with a budget line. The MCP connector assumes something else. A solo GTM engineer in Replit can call a `zoominfo.get_company()` tool the same way they call `stripe.charge()`, and licensing gets sorted out later. ZoomInfo is the first B2B data company to show in shipped code that reaching developers now matters more than selling seats. Clay, Apollo, and LeadIQ have every reason to ship their own MCP endpoints before the end of Q4.

Google's 'Call for Me' is the most consumer-focused of the four, but the effect is the same. Every SMB voice vendor (HubSpot's inbound-call product, Podium, Numa, Grasshopper, Ringly) assumed a human would be on the other end of the call. Once Pixel 11 owners start routing reservation calls through Gemini, SMB voice vendors have a new buyer. It used to be the customer service manager comparing phone-menu features. Now it's the founder deciding whether their own agent can deal cleanly with a Gemini agent on the line. Google has tested the idea before. This week it moved the feature out of Search Labs and tied it to a phone Google sells. Hospitality, retail, and services vendors will feel this before Q1. The next pitch deck for an SMB voice startup will probably lead with how well it works with agents.

Grok Bot sets the price benchmark for the fourth surface, customer support. SpaceXAI claims a resolved ticket costs $0.20, against $1 to $4 per ticket for traditional AI support. eesel AI's breakdown of Grok Bot in support puts SpaceXAI's own internal figure at $0.30. Even at that level, the cost is about 3 to 13 times lower. The price matters less than it looks. Intercom Fin and Decagon made similar claims six months ago and got news coverage without changing the category. What matters more is who buys Grok Bot. It's usually an X-native founder who added it to the workflow because it already connects to Salesforce, HubSpot, Gong and now 1Password. The CS Ops director who would normally benchmark Fin against Ada is left working around a decision that has already been made.

Taken together, the four moves show more than agents getting better. Each vendor shipped a product this week that goes around a specific procurement step. Slack's B2B advantage was a buying decision coordinated by IT. Ando sells to a founder who wants a platform built for agents. ZoomInfo's advantage was a RevOps buying decision, and its MCP endpoint goes to an engineer who wants a quick lookup. Google's inbound-voice customer was a CX manager. 'Call for Me' serves a Pixel user whose call gets answered by whichever business handles agents well. Grok Bot's competitor was a CS Ops budget tier, but Grok Bot sells to a founder who has already built it into the workflow.

In practice, "agents as first-class users" describes a change in who signs the purchase order. It requires no claim about AGI.

The bear case worth taking seriously

Two arguments cut against this, and both deserve an answer.

The first is that seed rounds and MCP integrations don't prove adoption. Ando has $20M and no paying enterprise customers yet. ZoomInfo's Replit MCP could sit in the marketplace for a year without moving seat-license numbers. 'Call for Me' only works on Pixel phones, which make up under 4% of the US install base. Grok Bot's $0.20 figure comes from SpaceXAI and hasn't been independently tested. All four are product launches, and none proves that customers are buying. That's true, and it still leaves the argument standing. Where vendors are shipping is an input for Q4 planning, even before revenue catches up. Picture a CRO whose stack was built around IT-coordinated seat expansion in 2024. That CRO needs to know the direction now, because rebuilding the enterprise motion around founders and agents takes more than a quarter.

The second is that the "founder as buying committee" idea breaks down at scale. At a Series C company with 200 reps, the CRO still runs the RFP and the CIO still runs the security review. That's also true. But AI-native GTM vendors at around $50M ARR (Clay, Regie, 11x, Rox) sell into committees where the founder sits closer to the tool decision than in older categories. And this week's two CRO hires at Consensus and OneTrust show the enterprise motion catching up faster than most people noticed. The founder-as-buyer phase won't last. It's a window that opens when a category shift goes around the traditional gatekeepers, and it closes once they rebuild the gate.

What this means for CROs

Three moves, sequenced against the next planning cycle.

First, check how every AI-native GTM tool your company bought in 2026 actually got approved. If the founder or CTO put a tool into the workflow before the SOC 2 review closed, count it as a founder-buyer signal. Say those tools add up to more than 30% of the AI budget. Then your next Q4 conversation with the CFO should be about a framework for authorizing agents, and vendor consolidation can wait. This week's piece on agents that spend real money raised the same kind of question for ad tech. This is the sales-side version.

Second, ship an MCP endpoint for your product before Q1 planning. If you don't, explain in the board deck why your customers will be reached through a competitor's endpoint by year end. ZoomInfo likely went first because it has the most seat-license revenue at risk in B2B data. Every category has its own ZoomInfo (Clay in enrichment, Gong in call intelligence, HubSpot in CRM). Each will face a competitor with an MCP endpoint before Q1. Treat it as a distribution decision. It belongs in the same board discussion as yesterday's per-seat pricing piece, because per-seat pricing and closed integrations rest on the same 2024 assumption.

Third, if your company already has more than 50 registered agents in Slack, Teams or a homegrown tool, add an Ando pilot to the Q4 vendor list. Ando may not be ready this quarter. But the platform question of agent identity, permission scope and audit trails will be settled in the next 18 months. Testing a purpose-built platform against Slack and Slackbot should be part of that decision from the start. LinkedIn's identity push with Vouching and its AI-slop crackdown is doing the same work one layer up. Across the AI stack, companies are reorganizing around who or what has an identity.

The founder-as-buying-committee pattern won't stay this pronounced. IT will likely rebuild security around agent identity by 2027. Procurement will follow by 2028 with rules for agent authorization, and most categories will end up with a mix of founder and committee buying. My bet is that the next generation of AI-native GTM category leaders gets decided in the roughly two years before that rebuild. The vendors that shipped this week are backing that bet with product. Within 18 months the gate they went around will be rebuilt, so any vendor whose sales motion depends on skipping it has limited time.

In practice, "agents as first-class users" describes a change in who signs the purchase order.

Check how every AI-native GTM tool your company bought in 2026 got approved. If tools the founder or CTO put in before security review make up more than 30% of the AI budget, bring a framework for authorizing agents to your next CFO conversation.

By Business of GTM