That was the opening claim of our webinar, The AI-First Customer Org, with Gal Biran (Co-founder and CEO, Base AI) and Shachar Avrahami (VP Product, Base AI). One hour. Five customer programs shown live, with the real numbers behind them. The session draws on what we see running 500+ customer programs that touch more than 2 million B2B customers every month, for organizations like SAP, Okta, Broadcom, and ZoomInfo.
This recap covers the full argument, the five use cases, and the numbers. If you would rather watch it happen on screen, the recording is here.
Why AI-first, why now
Your customers already made the shift. More than half of customers say they do not care whether a person or an AI solves their problem, as long as it gets solved. Over 80% of leaders agree that GenAI can provide better service than previous approaches. When customer preference and leadership preference point the same way, the org changes whether you are on board or not.
Three forces compound it:
Smaller teams, more ownership. With an average reduction in force around 25%, each person owns more. Tools consolidate. Roles consolidate. The question shifts from headcount to system design.
Customer 360 became non-negotiable. Everyone talked about it for years. AI made it real, because your post-sale AI is only as smart as the context behind it. Product telemetry, CRM, warehouse, support, all in one place, or the gaps become risk.
The pace tripled. What took a quarter is now expected in a month. Crawl, walk, run is over. The new motion is launch and iterate.
And one more that gets underestimated: AI already meets your customers before you do. A support agent, a website agent, an onboarding flow. The first experience most customers have with your company is an AI experience. The only question is whether it is a designed one. If you want to pressure-test where your own org stands, bring your questions to a session with us.
What an AI Engagement OS actually means
Base is the AI Engagement OS for customer-led growth. In the webinar, that phrase stops being positioning and becomes an architecture. It means one AI system for all customer programs, built from two kinds of AI working on one shared context layer:
AI agents own the actions. The workflows, the processes, the jobs to be done. A handoff agent that reads your Gong calls, closed-won notes, and CRM data, then briefs CS automatically. An onboarding agent that schedules the kickoff, personalizes the journey, and alerts when an account goes off track.
Vibe-coded experiences own what the customer sees. Onboarding journeys, QBR hubs, communities, loyalty programs. Designed with AI, personalized per account and persona, and embedded where customers already are: your product, your portal, your community.
A great AI-first org uses both. Agents are the engine, experiences are the interface, and the shared context layer underneath connects your CRM, product usage, support history, and warehouse so every agent and every experience works from the same truth.
Gal set one bar for all of it. Whatever you build has to be relevant to the customer’s stage and persona, easy to access where they already work, human in how it communicates, and actionable enough that the next step is obvious. Miss one of those and the smartest agent in the world reads as noise.
Map the 50 jobs to be done before you build anything
The framework that anchors the session: most organizations have 50 or more post-sale jobs to be done. Getting integrations connected. Booking the kickoff. Running the feedback survey. Getting a G2 review. Preparing the business review. Each one maps to a stage in the customer lifecycle, a persona, and a product line.
For every job, ask one question: does this need a human, AI with a human in the loop, or full autonomy?
The mapping takes a couple of days. It is the highest-leverage planning exercise in the whole approach, because it turns “we need AI” into a prioritized backlog: the outcomes that are most critical, or easiest to move to AI, go first. Launch one, iterate, add the next. Being AI-first does not mean AI everything. It means running this question on every new process and every old one, then automating the ones that earn it.
Five programs, shown live
Shachar’s half of the session is the part you should watch rather than read. Five real systems, from real customer organizations, with his live builds in between. The recording shows all of them end to end, and each section below has its own short cut.
1. Self-serve onboarding: ZoomInfo
ZoomInfo built fully self-service, personalized onboarding. Customers launch their own journey, set their own goals, timelines, and integrations, and move through a gamified experience that celebrates milestones along the way. No rep required, full visibility for the customer at every phase.
The numbers: a 100% self-service flow, kickoffs 2x faster, and implementation cut from weeks to days, with CS headcount reallocated to higher-value work. Mid-section, Shachar vibe-codes a working onboarding program live, straight from a Zendesk help center, and reviews the AI-generated draft on screen.
2. Autonomous success plans: Okta
Post-onboarding, most self-serve customers have nobody to talk to. Okta built autonomous success plans for thousands of self-service customers with a four-step recipe: capture customer goals and log them to CRM, calculate each account’s maturity and missing configuration from live product data, generate a personalized plan with prioritized tasks, and auto-check tasks as customers complete them in the product. Hit a threshold and more advanced setup unlocks.
The quiet insight: nobody had ever asked those customers what they wanted to achieve. Asking, then acting on the answer, is a retention motion on its own.
3. Always-on QBRs: Vidyard
QBR week is screenshots, manual decks, and coverage for top accounts only. Vidyard replaced it with an always-on, AI-powered QBR embedded in their product, built on live telemetry. Every customer gets a real-time value snapshot they can act on themselves.
The numbers: 100% QBR coverage, every customer and every segment, with onboarding 8x faster (from around 2 months to around 1 week) and in-hub engagement rising quarter over quarter.
4. AI Documents: one design, a personalized PDF per account
Launched in this session: AI Documents, a new module you can also get standalone. Design a QBR, renewal review, presentation, or one-pager once, and generate a personalized PDF per account with data pulled in real time from Salesforce, Snowflake, BigQuery, Databricks, and the rest of your stack. Customers can generate their own value snapshots on demand.
The renewal version writes itself: connect the renewal opportunity in Salesforce, and 90 days out the account gets a renewal overview with an auto-renew ask built in. If you want to see AI Documents running on your own accounts, book 30 minutes.
5. References, run by four agents
Every CS team gets the same ask from sales: we need a reference to close this deal. The webinar maps a reference program run end to end by four agents. A recruiting agent scans call recordings, support history, and product usage to find likely advocates and reaches out. A matching agent takes each incoming request and picks the right reference for the deal, handles consent, and makes the intro. A scheduling agent finds the overlap and gets the meeting booked. A content agent arms the rep with case studies and quotes while the match is in flight.
Reference automation like this influenced more than $50M in pipeline, with a 100% self-serve flow, 260 AI-generated reviews, and 100+ AI-built customer stories. Advocacy stops being a favor economy and becomes a system.
The metrics change too
When CS owns more of the journey and all of it runs on one system, the reporting conversation changes. NPS, CSAT, and health scores stay useful. But the board conversation moves to pipeline influenced, net revenue retention, advocates secured, and deals accelerated by references.
Gal’s closing argument goes one step further. When anyone can vibe-code a product, features stop differentiating. What remains is service, know-how, and experience. That makes customer experience the moat, and it makes this a very good time to work in customer success. This is the same shift we wrote about in From Customer Success to Customer-Led Growth, and it lands hardest when CS and customer marketing operate as one system. ZoomInfo already runs that way: onboarding, references, advocacy, and customer marketing all live under one Digital Success and Scaled CS org.
Three moves to start on Monday
1. Run the jobs-to-be-done mapping
Block two days. List your post-sale outcomes, all 50+, and tag each one: human, AI with human in the loop, or fully autonomous. Rank by criticality and ease. This backlog is your AI roadmap, and it costs nothing but focus.
2. Launch one program and iterate
Pick the motion where you are most manual: onboarding, business reviews, or references. Ship the smallest version that touches real customers, then improve it weekly. Launch and iterate beats crawl, walk, run. If you want a shortcut, we will build the first one with you on your data.
3. Put revenue numbers on it from day one
Instrument pipeline influenced, NRR impact, and time-to-value before you launch, not after. The teams in this webinar can defend their programs to the board because the numbers were built in: $50M+ pipeline influenced, 100% QBR coverage, onboarding from weeks to days.
Systems beat headcount. The CS teams that win the next three years will have the best systems, not the biggest teams.
An AI Engagement OS is two kinds of AI on one context layer. Agents own the actions, vibe-coded experiences own what customers see, and both work from the same data.
Map your 50+ post-sale jobs to be done first. Tag each one human, AI with human in the loop, or autonomous, then launch and iterate.
The numbers are real. ZoomInfo: 100% self-serve onboarding and $50M+ pipeline influenced. Vidyard: 100% QBR coverage and 8x faster onboarding. Okta: autonomous plans for thousands of customers.
Report revenue, not just health. Pipeline influenced, NRR, and deals accelerated by references are the new CS board metrics.
