Our first AI product reached $100M in ACV within six months of launch. Most of what made that possible happened before and after launch day, and almost all of it is repeatable.
This play breaks the launch into 10 steps. Steps 1 to 4 cover the quarter before launch; steps 5 to 8 cover the launch and the sales motion. Steps 9 and 10 cover what happens after the sale. Each step shows what we did and what you can do yourself when launching a product of your own.
Our AI product went live in May 2024. Within 6 months it reached $100m in ACV. Within 18 months, it reached $250M, including customer upsells.
1. Make your team customer zero
Before customers see it, your own GTM team uses the product as their main daily tool. Their usage is the beta, and their stories become the pitch.
What we did
1,600 of our own GTM professionals used it first, and we reshaped the product around who used it and how heavily.
Our GTM team spent Q1 2024 using it to refine talk tracks and test collateral.We later tied rep confidence directly to this.
For the first six months, the demo was a day in a rep’s life: AI pushing the reason to call, from an earnings-call comment or a new funding round, with a message drafted from those signals.
Your move. Make sure to run internal use for a full quarter, and build the launch story and the demo around the one or two workflows your team uses most. It only works if your team does the buyer’s job.
2. Sell it before launch
Pitching and selling before general availability tests demand, shows what’s missing and collects the language buyers use.
What we did
We pitched it ourselves to dozens of customers. Upper mid-market and enterprise buyers kept saying they’d been trying to build something similar for years, which became a core value argument.
We had two small GTM teams sell it in live deals with real prices, and sorted every missing feature into a deal-breaker or a planned follow-on.
Your move. Book 20 to 30 exec-led pitches before launch, give one or two teams a draft price card, and log every objection and missing feature.
3. Prove it in beta
Run the beta with the launch statistics planned from day one, including a comparison group.
What we did
More than 20,000 people used the beta before general availability. We ran it inside existing customer accounts, so beta users could be measured against colleagues in the same roles who weren’t using it.
They reported about 10 hours a week saved on research, and signals behind 45% of opportunities created.
Beta users created nearly twice as many opportunities as non-users in the same roles at the same companies. That comparison makes the number hard to dismiss.
Your move. Decide your launch metrics and comparison groups before the beta starts.
4. Enable before you sell
Train, certify and equip every seller before customers start asking questions, with a daily loop back to leadership.
What we did
A virtual onboarding program for sellers, with certifications, pitch decks, demos and a 20-page Success Guide.
Leadership met daily with the whole GTM team to go through wins, challenges and learnings.
Your move. Certify every customer-facing role before launch, ship a Success Guide as standard, and hold a daily launch stand-up for the first quarter.
5. Launch to the accounts most likely to get value
Score customers on how likely they are to get value, launch to the top tier first, and match the team behind them to that segment.
What we did
Our first rollout targeted customers by usage, integrations and ICP fit. CRM integration was the hard filter, because the product’s value comes from combining a customer’s CRM data with ours.
Over 75% of early upsells were mid-market or enterprise.
Enterprise reps handed to enterprise onboarding, CSMs and AMs. Segmenting new business reps by size produced our best mid-market and enterprise new business quarter.
A quarterly review of every product in every segment, on retention and growth, decides where we invest, fix or partner.
Your move. Find what your product depends on (for us, CRM integration), filter your launch list on it, and put an upmarket delivery chain behind an upmarket launch.
6. Launch where the market looks
Announce early, backed by customers and partners your buyers already trust, then keep the proof fresh with every release.
What we did
The announcement came with a launch video, an early-adopter customer quote and an endorsement from Anthropic’s co-founder, and was paired days later with our earnings release.
Every seasonal release brought new proof. The summer release reported nearly 25% more pipeline.
Our customer impact report put users at 10.5 hours saved a week. An ROI study we commissioned from GTM Partners reported $2M in new revenue, 2x conversion and 5x productivity.
We later changed our Nasdaq ticker from ZI to GTM, alongside the GTM Studio launch.
Your move. Use the gap between announcement and launch to build pipeline, plan a release about every quarter, and make the category claim only once the product proof is in.
7. Run three revenue routes, and hold the line on price
New logos, off-cycle upsells and migrations at renewal each need their own owner, and existing customers should pay for the extra value.
What we did
All three routes ran in parallel through 2024. In the back half of 2025, 98% of new business customers came on to the product.
Migrations came with double-digit uplift. Where a customer had fewer users, a flat renewal in place of a potential downsell counted as a win.
Packaging changed with what we learned from the beta and focus groups, and AI became part of the main tiers: 'Advanced' and 'Enterprise'.
Your move. Give each route its own target, set an outcome and a discount floor for each type of account, and decide upfront who gets AI bundled and who pays for it.
8. Grow accounts team by team
Lower the risk of buying with team-level trials, then grow by adding new roles and replacing other vendors.
What we did
Customers could trial it team by team or user by user. One 100-seat proof of concept at a large job search engine grew more than tenfold.
It moved beyond SDRs into AEs, AMs and CSMs, a user base more than three times larger.
A mid-market tech customer with a mandate to cut spend by 20% ended up with a 60% ACV increase by consolidating vendors onto ZoomInfo.
Your move. Agree a trial’s success metrics and scale-up plan upfront, map each new persona to one use case, and give AMs a vendor-replacement case for tight budgets.
9. Get Customer Success ready on day one
Plan post-sale capacity from the sales forecast and train CS before the deals land, so onboarding never becomes the bottleneck.
What we did
31 technical deep-dive sessions for CS and CX.
Used the sales forecast to work out how much onboarding work was coming each month, and how many people we’d need to handle it alongside existing customers.
At onboarding, AI reads the customer’s website, case studies and press releases to set the product up around their market.
Daily signals pushed through email, Slack, Teams and mobile brought under-used seats back. Customers on it engaged about 40% more, and the first renewal group renewed meaningfully better than legacy customers.
Your move. Use your sales forecast to plan onboarding resource before launch, automate what you can, and point CS at the lowest-usage seats first.
10. Keep shipping after launch
Launch day is the start. Ship fast from field feedback, and let real usage decide what gets built next.
What we did
We shipped 130+ features in the six months after launch, with 64 engineers working only on signals.
Real-time usage data across our own reps and customers decides where investment goes. Account AI’s 50% usage increase earned it more.
Our number one rep metric is signals actioned: whether reps act on the buying signals in their accounts, and how good the outreach and first call were.
Your move. Instrument usage from day one, review it alongside revenue when you decide what to build, and make signals actioned the headline rep metric.
FAQs
How long should internal use run before launch?
A full quarter. That’s long enough to see which one or two workflows your team uses most, and those become the launch story. It only works if your team does the same job as your buyers.
What should an AI product beta be designed to prove?
Decide the launch metrics and comparison groups before it starts. Our beta compared users with non-users in the same roles at the same companies, which made the results harder to dismiss.
Who should get a new AI product first?
The customers most likely to get value from it. Find what the product depends on (for us, CRM integration) and launch to the accounts that already have it.
How do you price an AI upgrade for existing customers?
Charge for the extra value rather than giving it away to secure renewals. Set a target outcome for each type of account, from uplift to a flat renewal in place of a potential downsell, and a floor on discounting.
How do you stop post-sale becoming the bottleneck?
Use the sales forecast to work out how much onboarding work is coming each month, and how many people you’ll need, before the deals land. Then automate as much of onboarding as you can.
What one metric should reps be measured on?
Signals actioned: whether reps act on the buying signals in their accounts, with outreach quality and first-call quality underneath.