Of the 28 companies in our portfolio that have raised seed rounds, 22 have either already raised Series A funding or are actively in the process of doing so. That is a 94 percent follow-on rate -- significantly above industry averages -- and it is not an accident. It reflects both our selectivity at the seed stage and the systematic support we provide to portfolio companies on the path from seed to Series A.
Over the past three years, we have had front-row seats to what separates AI companies that close successful Series A rounds from those that do not. This article is our attempt to synthesize those observations into a practical playbook for founders who are currently in seed mode and thinking about what they need to accomplish before entering the Series A fundraising process.
Fair warning: this is going to be more honest and direct than most VC writing about fundraising, because we think founders deserve accurate information rather than comfortable platitudes. The bar for Series A has risen significantly in the past 18 months, and many AI founders who expected their technical achievements to carry them to easy follow-on funding are discovering that the process is more demanding than they anticipated.
The Core Insight: Series A Is a Revenue Conversation
The single most important thing we have learned from watching our portfolio companies raise Series A rounds is this: Series A is fundamentally a revenue conversation, even for AI infrastructure companies. This is a significant change from the 2021-2022 era, when AI infrastructure companies could raise large Series A rounds on the basis of developer adoption, GitHub stars, and an impressive technical team without meaningful revenue.
That era is over. The top-tier Series A funds we work with -- Sequoia, Andreessen Horowitz, Benchmark, Thrive, and others -- are now applying much more rigorous revenue filters to their decisions. For SaaS-model AI companies, they want to see $1M to $3M ARR for an early Series A ($15M-$25M) or $3M to $6M ARR for a larger round. For infrastructure and developer tool companies, the revenue bar can be lower if there is an exceptionally compelling usage and growth story, but even then, $500K ARR with clear enterprise pipeline is a minimum expectation at most tier-one firms.
The good news for founders in our portfolio is that we knew this coming into seed, and we structured our portfolio companies' seed plans accordingly. Eighteen months of runway, a focused go-to-market motion aimed at the 10 enterprise customer segments most likely to generate early revenue, and a clear definition of what "Series A ready" looks like before we wrote the seed check. If you do not have that clarity yet, now is the time to establish it.
The Four Metrics That Series A Investors Actually Evaluate
Beyond the ARR threshold, there are four specific metrics that we have observed tier-one Series A investors spend the most time on when evaluating AI companies. Understanding these metrics -- and building your business to optimize for them -- is the core of the seed-to-Series-A playbook.
1. Net Revenue Retention (NRR) Above 120%
Net Revenue Retention measures how much revenue a company retains from its existing customer base over time, including expansion revenue. For AI companies specifically, NRR above 120% is a powerful signal that (a) the product is genuinely driving value for customers, (b) that value is expanding as customers use the product more deeply, and (c) the product has the kind of workflow integration that makes churn unlikely.
NRR is particularly important for AI companies because it addresses one of the most common concerns investors have about AI products: that customers will try them, find them impressive but not indispensable, and then cancel at renewal. High NRR directly refutes this concern. Every portfolio company in our portfolio that has successfully raised Series A had NRR above 115% at the time of their fundraise. The companies that struggled had NRR in the 90-105% range -- technically not churning, but not demonstrating the expansion revenue that signals product-market fit.
2. Revenue Per Customer Above $50K ARR (for Enterprise)
The size of your contracts matters enormously for Series A. Ten customers paying $100K ARR each is a dramatically better business than 100 customers paying $10K ARR each, even if total ARR is the same. Higher contract values signal that you are solving a genuinely critical enterprise problem that customers are willing to pay meaningfully for. They also suggest that you can scale to significant revenue without an enormous sales headcount, which matters enormously for gross margin and operating efficiency projections.
For enterprise AI companies, we counsel founders to resist the temptation to close many small deals early in favor of fewer, larger ones. A single $100K contract with a Fortune 500 company that proves the use case, generates compelling case study material, and provides a strong reference is worth more for your Series A than five $20K contracts with mid-market customers who are often less demanding and therefore less reliable signals of true product-market fit.
3. Payback Period Under 18 Months
Payback period -- the time it takes for a new customer's revenue to cover the cost of acquiring them -- is a proxy for capital efficiency and sales model quality. Series A investors want to see that your AI company can grow efficiently, not just that it can grow. A payback period under 18 months demonstrates that your sales motion is working, your average contract value is appropriate for the effort required to close deals, and your unit economics support continued investment in growth.
We have seen AI companies with impressive ARR growth and terrible payback periods -- often because they were subsidizing sales with heavy engineering support, extensive proof-of-concepts, and long implementation cycles. These businesses look strong in ARR terms but are actually burning cash at an alarming rate. Smart Series A investors will find this, and it will either kill the deal or significantly reduce the valuation.
4. Gross Margin Above 65%
AI companies have a gross margin problem that traditional SaaS does not: inference costs. Every time your application calls a foundation model API, that call costs money -- and at scale, those costs can consume a frightening fraction of your revenue. Companies that built their products assuming that model inference would always be cheap are discovering that their gross margins at scale are dramatically lower than the 80%+ that software investors expect.
The solution is to either fine-tune smaller, cheaper models for your specific use case (reducing inference cost by 10-100x compared to frontier models), build caching and batching infrastructure to reduce redundant inference calls, or price your product in a way that passes inference costs through to customers. Whatever approach you take, you need to have a credible path to 65%+ gross margins before entering the Series A process, because software investors will do the math and the conversation about inference costs will happen whether you address it proactively or not.
The Narrative That Opens Doors at Tier-One Funds
Getting the metrics right is necessary but not sufficient for a strong Series A fundraise. You also need a compelling narrative that contextualizes those metrics within a larger market thesis that resonates with the specific investors you are meeting with. After dozens of Series A fundraise post-mortems with our portfolio companies, we have identified three narrative frameworks that consistently work with top-tier funds.
The "category creation" narrative argues that your company is defining an entirely new software category that did not exist before your specific AI capability became available. This narrative works when you can demonstrate strong customer pull even in the absence of an established buying category -- customers are paying for something they could not even have articulated as a need 18 months ago. DataNexus used this narrative effectively by arguing they were creating the "semantic search infrastructure" category, a category that simply did not exist before enterprise LLM deployment made vector search a necessity.
The "wedge into a large market" narrative argues that your current product is a beachhead in a much larger addressable market. This works when your current product is demonstrably winning in a specific segment, and there is a clear expansion path into adjacent segments over the next 24-36 months. Cognex Health used this effectively by showing strong performance in radiology and a credible expansion roadmap into pathology, oncology, and eventually general clinical decision support.
The "workflow transformation" narrative argues that your product is not just improving an existing workflow but fundamentally replacing it with a more efficient AI-native workflow. This works particularly well for enterprise AI companies that can demonstrate measurable productivity improvements -- if your customers are doing a task in two hours that previously took two weeks, that is a narrative that practically sells itself.
The Timeline: What to Do and When
Based on our experience with portfolio companies, here is the 18-month timeline we use to help seed-stage AI companies prepare for a strong Series A fundraise.
Months one through six are about finding product-market fit. Your only goal in this period is to identify the customer segment and use case where your AI product delivers undeniable, measurable value. Deploy to 5-10 design partners willing to provide detailed feedback and serve as references. Track usage and outcomes obsessively. Do not worry about scaling sales -- worry about finding the insight that will make scaling possible.
Months seven through twelve are about proving the revenue model. Convert your design partners to paying customers and sign your first two to five arm's-length enterprise contracts. Start measuring NRR, payback period, and gross margin. Hire your first sales or customer success leader. Build the internal reporting infrastructure that will make your metrics legible to Series A investors.
Months thirteen through sixteen are about generating momentum. Start a deliberate outbound sales motion targeting the 20 highest-probability enterprise customers in your target segment. Aim to double ARR. Build a pipeline of 10-15 qualified opportunities. Start meeting informally with Series A investors to gather feedback and build relationships without creating formal fundraising pressure.
Months seventeen through eighteen are the fundraise. With $2-4M ARR, strong NRR, and 5-10 enterprise referenceable customers, you are in a strong position to run a competitive process with 8-12 top-tier Series A funds simultaneously. Use your seed investors' networks to get warm introductions to the specific partners at each fund who are most likely to champion your space. Move quickly -- the best Series A processes close in 6-8 weeks from first meeting to term sheet.
The Mistakes That Kill Series A Fundraises
We have also observed patterns in companies that struggled with their Series A, and it is worth naming the most common mistakes. Premature scaling -- hiring a large sales team before product-market fit is proven -- is the most common and most damaging error. Companies that hire aggressively before they have a repeatable sales motion end up with high burn rates, low ARR relative to headcount, and metrics that tell a story of inefficiency that is very difficult to overcome in a fundraising process.
Chasing vanity metrics is the second most common error. GitHub stars, API call volume, developer signups, and NPS scores are interesting data points, but they do not substitute for revenue at the Series A stage. Founders who focus on these metrics at the expense of revenue development often arrive at their Series A process with impressive vanity metrics and disappointing ARR, which creates a narrative problem that is difficult to resolve.
Poor investor relationship management during the seed period is a third common error that founders consistently underestimate. The Series A investors you will approach in 18 months have a much stronger signal on your company if they have been watching it quarterly for 12 months. Start building these relationships early -- share monthly updates with your target Series A investors, invite them to customer calls and product demos, and ask for feedback that you can demonstrate acting on. By the time you formally start your Series A process, the best investors should already have a mental model of your trajectory that makes diligence faster and conviction higher.
Our Commitment to Portfolio Companies on This Journey
At Milestone AI Ventures, helping our portfolio companies navigate the seed-to-Series A journey is one of the highest-value activities we engage in. We have established relationships with the relevant partners at 30+ top-tier Series A funds who actively source from our portfolio and trust our judgment on the companies we have backed. We provide explicit Series A preparation programming starting at month 12 for each portfolio company, including metrics benchmarking against our full portfolio and introductions to Series A investors calibrated to each company's specific stage and sector.
The path from seed to Series A is not easy -- it requires genuine product-market fit, disciplined execution, and a company that is demonstrably more valuable eighteen months after we invest than it was on the day we wrote the check. But for the AI companies that get this right, the current environment offers some of the best Series A conditions we have seen in years. Capital is available, valuations are reasonable, and the enterprises buying AI solutions are doing so with increasing urgency. The opportunity is real -- and we are here to help our portfolio companies capture it.
Marcus Rivera is a General Partner and Co-Founder of Milestone AI Ventures. He previously served as CTO at Databricks and Senior Principal Engineer at Palantir. The views expressed here are his own and do not constitute investment advice.