# How AI Agents Let GTM Teams Scale Justin Joyce, Cloudflare | AI Engineer World's Fair 2026 | 19:15 Source: https://www.youtube.com/watch?v=Qw_tC68KKes Channel: AI Engineer (https://www.youtube.com/@aiDotEngineer). Summarised by AIE Talks. Page: https://aietalks.com/talks/how-ai-agents-let-gtm-teams-scale Published: 2026-08-26 Tags: agents, enterprise, multi-agent, workflows ## TL;DR - Traditional go-to-market work does not scale because operations teams repeat analysis manually while salespeople lose time gathering context and lack the experience of top performers. - Cloudflare uses business-aware skill files, pushed summaries, and self-service agent workspaces to give teams faster access to analysis and expert guidance. - A three-agent workflow drafts, checks, and rewrites weekly summaries, while curation and feedback loops make the system more predictable. ## Summary Justin Joyce argues that traditional go-to-market operations break down as teams grow. Operations staff rebuild analysis in spreadsheets or maintain dashboards that cannot answer every request. Salespeople also lose time switching between prospects, customers, and adoption work while trying to gather the right context. Joyce describes three responses at Cloudflare: scale analysis with skill files that combine business knowledge and data logic, push insights to teams instead of making them search dashboards, and provide self-service tools for forecast briefs, QBR decks, account plans, and renewal preparation. A weekly summary is produced by a drafter, a reviewer, and a tone agent, with each run inspected for two to three months. Cloudflare OS gives go-to-market users an agentic workspace connected to curated skills and company data. Joyce reports twice the efficiency, while admitting that CRM updates, quoting, approvals, security, and controlling the growth of skills remain harder problems. ## Key ideas ### Traditional go-to-market work grows faster than operations capacity [01:59](https://www.youtube.com/watch?v=Qw_tC68KKes&t=119s) Joyce says traditional go-to-market does not scale because operations teams often rebuild analysis in Excel or Sheets every week. Each additional project adds more repeated work. Dashboards help leadership and executives, but they meet the needs of only some users. As a result, go-to-market teams still cannot get every piece of information when they need it. Joyce's goal at Cloudflare is to make the operation more efficient from the back office through to the sales team, using the machine learning and sales operations experience he developed across his career. ### Salespeople lose time to context switching and uneven experience [03:01](https://www.youtube.com/watch?v=Qw_tC68KKes&t=181s) Joyce describes two gaps on the sales side. The context gap appears when someone moves from a prospect call to a current-customer call and then to an adoption conversation. The salesperson must gather different information between each interaction. The expert gap is the difference between how an experienced seller handles a prospect, adoption issue, or customer satisfaction problem and how a new salesperson handles it while ramping up. Joyce wants teams to work with more consistent execution and messaging, including a shared way to assess customer problems and connect them to Cloudflare's products. ### Cloudflare's approach has three connected pillars [04:56](https://www.youtube.com/watch?v=Qw_tC68KKes&t=296s) The first pillar is scaling analysis, so operations can answer data questions and build applications with business context. Joyce gives an example of reducing work that took two hours to five minutes. The second pillar is scaling insight by sending the story in the data to teams at weekly and management-specific levels, including the customers salespeople discuss. The third pillar is self-service, so sellers can get expert-level information while handling a customer, including guidance on approach, rejection, upsell opportunities, and satisfaction issues. ### Skill files give nontechnical users business-aware access to data [06:53](https://www.youtube.com/watch?v=Qw_tC68KKes&t=413s) Cloudflare builds role-specific skill files that connect business information to the data. They support technical users who write SQL and business users who do not. The files include questions the business commonly asks, such as why opportunity close dates changed or why opportunity amounts changed. Joyce says this covers 80% or more of routine questions, leaving more complex strategic questions for other work. Users who previously had to wait for someone who knew SQL can ask the data questions directly. The same business and table information also helps the team build applications faster. ### Insights should reach people before they open a dashboard [09:00](https://www.youtube.com/watch?v=Qw_tC68KKes&t=540s) Joyce says there is a story in the data and people should not have to search for it. Cloudflare sends a weekly summary covering business performance, progress toward goals, trends, standouts, and watch items. Dashboards still have a place for standard information and deeper investigation, but people adopt KPI metrics unevenly. A pushed summary gives less dashboard-oriented users a regular view of what is happening, while allowing others to drill into reports when they need more detail. Cloudflare prepares the data by time, business slice, and metric before agents analyze it. ### Three agents separate drafting, data review, and tone [11:18](https://www.youtube.com/watch?v=Qw_tC68KKes&t=678s) The weekly summary uses a multi-agent workflow. The first agent gets the data and creates a draft, calling the team's MCPs. A second reviewer checks whether the draft is supported by the data. A third tone agent rewrites the message so risks and opportunities receive equal weight. Joyce says the team has observability into each LLM call, including what was passed in and what response came back. They watched every run for about two to three months before deciding that this setup was working for them. ### Cloudflare OS puts curated expert skills in sellers' hands [12:09](https://www.youtube.com/watch?v=Qw_tC68KKes&t=729s) Cloudflare OS is an internal agentic workspace running on Cloudflare. It gives go-to-market users their own compute and persistent environment through Cloudflare Workers and Durable Objects. Users apply it to forecast briefs, QBR decks, purchase decks, account planning, general data questions, and renewal preparation. The workspace connects to expert skills, an MCP connection, and the AI Gateway. A central skill repository is curated and reviewed by go-to-market and operations teams so skills do not proliferate without control. Joyce shows the system generating daily work plans and a custom slide deck for a customer call. ### Curation and feedback make agent work more predictable [15:43](https://www.youtube.com/watch?v=Qw_tC68KKes&t=943s) Joyce calls skill curation the basis of the agentic workforce. Business knowledge and analyst skills give agents the information needed to answer questions, build applications, and guide customer work more consistently. He also stresses an internal feedback loop: teams need to see whether what they built is useful and where users have problems. The three pillars work together because some users ask operations directly, some need information pushed to them, and others want to use a self-service workspace. Joyce reports that this combined approach has doubled the team's efficiency. ### CRM writes and governance are still unfinished [17:11](https://www.youtube.com/watch?v=Qw_tC68KKes&t=1031s) Joyce expects deeper connections to the systems where go-to-market work happens. QBR decks and renewal-call skills could be attached to meetings automatically, and meeting notes could be collected across the relevant systems. He says quoting, approvals, and updating Salesforce are harder tasks that require carefully designed workflows and security. Cloudflare is building those connections. Joyce also describes the current period as a Cambrian stage, with an explosion of agent skills and experiments. As the systems become more integrated, teams will need a strategic way to control how each group uses them while keeping sources of truth aligned. ## Notable quotes - "The general problem is that traditional go-to-market does not scale." (01:59) - "There's story in the data, and they really shouldn't have to search for it." (08:50) - "The first thing is skill curation is the basis for all of this agentic workforce." (15:40) - "Through all these different pillars that I talked about, we've really been able to 2x our efficiency." (16:56) ## Tools & references mentioned - Cloudflare - Cloudflare OS - Cloudflare Workers - Durable Objects - Salesforce - MCPs - AI Gateway - Gemini ## Who should watch - You run sales operations or revenue operations and spend each week rebuilding reports, answering repeated data questions, or maintaining dashboards that do not cover every team need. - Your sales team loses time preparing for calls, account reviews, renewals, or QBRs because useful context is spread across systems and the difference between experienced and new sellers is large. - You are introducing internal agents and need practical guidance on skill curation, staged testing, feedback loops, data review, and the limits around CRM writes and approvals.