Revenue Builders · 17 Sep 2026 · From the week of 14 September
Why Proprietary Data Beats AI Software for 2x Pipeline Growth with Ann Davis
These are notes on the conversation, checked against its transcript. The episode itself has the full discussion.
In brief
Ann Davis, Chief Revenue Officer at Crunchbase, joins hosts John Kaplan and John McMahon to argue that proprietary data matters more than software in the AI era. She describes Crunchbase's private-market dataset, its MCP server connecting AI tools to Salesforce and Gong, and what she changed when moving the company from product-led growth to enterprise sales. Her central point is that AI is only as good as the data it is built on, so data providers should build datasets competitors cannot replicate and deliver them inside the seller's workflow.
For founders
- Build proprietary datasets that competitors cannot replicate, and deliver them through an AI interface inside the user's workflow so the product is not reduced to a system of record.
- Consider licensing data within software with a usage component, such as per-query pricing, because companies rarely buy enterprise-wide licences for data sold by seat or employee count.
- When moving from product-led growth to enterprise, set the board's expectations early that enterprise deals take 12 to 18 months, since directors tend to lose sight of why the timeline is long.
- Before taking on an enterprise pivot, Ann checks whether the company has mapped how AI could enhance or hurt its business over five years and whether the executive team is energised about the change.
For revenue leaders
- Ann says she pitches heads of sales that telling sellers which companies to pursue and which to avoid, using Crunchbase's data, drives 2x pipeline growth in six months and 25% top-line growth in the same period; she did not tie the figures to a specific customer result.
- Ann says an SMB deal takes about three months while an enterprise deal takes 12 to 18, and that demand generation without a budgeted project adds at least six more months.
- Ann says about 90% of the sellers she inherited opted out of the enterprise motion when given a window to prove enterprise skills, saying enterprise selling was not in their DNA.
- Ann says she has hired and let go two RevOps people who did not do the analysis she needed, and that she now handles much of that analysis herself through an AI front end on Salesforce and Gong.
- Ann says sellers must stay in the driver's seat of the conversation, because customers who steer the deal will take the seller down windy roads to a dead end.
What was said 29, most useful first
Telling reps which companies to pursue and which to avoid, using private-market data, is claimed to drive 2x pipeline growth in six months and 25% top-line growth in the same period. Listen
Ann says this is the consultative pitch she makes to heads of sales and CROs about Crunchbase's data, and that the edge comes from proprietary data rather than leaving reps to interpret AI output on their own. She presents these figures as the outcome the approach delivers. The company and stage behind the figures are not specified.
“drives pipeline growth 2x in six months and drives top line revenue growth 25 % in that same six months”
Listen to the episode Pipeline & demand generation Link to this
Data creates the most value when it is delivered at the workflow level through a headless AI interface such as an MCP server, not as a standalone dashboard. Listen
Ann says Crunchbase launched its MCP server about a month before the recording and that the value lies in bringing data sources together through an AI interface that sits inside the seller's workflow. She says the choice of front-end AI tool, such as ChatGPT, Gemini or Grok, does not matter.
“it's about how you seamlessly bring together the data sources through the AI interface to deliver real value.”
A daily rundown sheet connected to Gong, Salesforce and an MCP server can tell each rep what to do that day based on what changed since they last worked. Listen
Ann says her team built this sheet in about five minutes, and that it shows every sales rep what they need to do that day. The sheet draws on Gong, Salesforce and Crunchbase's MCP connection.
“my people spent five minutes and built a daily rundown sheet that's going to connect to Gong, it's going to connect to Salesforce, it connects to our MCP, and it lets every sales rep know exactly what they have to do in that particular day because of what happened since they were last working.”
To avoid being reduced to a system of record, a data provider should build proprietary datasets competitors cannot replicate and expand its data distribution partnerships. Listen
Ann says Crunchbase recognised before she joined that AI would likely take a large part of its business, so it prioritised datasets that competitors could not copy. She says the second part is supporting the new solutions being built on top of its data through a data distribution partnership strategy.
“the best way to do that is to develop proprietary data sets that they can't replicate.”
She could feed a 350,000-account book into Crunchbase and get a territory plan in five minutes, versus the five weeks her Google team spent building plans manually. Listen
Ann says that when she managed a team of 10 sellers at Google, the team spent January into February building territory plans by hand. She says she could feed in her 350,000-account book of business and get a territory plan back in five minutes using Crunchbase, and that as a seller she would have paid for that time saving herself.
“I could feed in my book of business of 350 ,000 accounts and get a territory plan back in five minutes with Crunchbase.”
When Crunchbase moved from product-led to enterprise, it gave sellers a window to prove enterprise skills, and about 90% of them opted out. Listen
Ann says she restructured the sales team early and quickly because she had numbers to hit. Most people who were given the chance to show enterprise skills chose to leave the enterprise motion, saying enterprise selling was not in their DNA.
“to be honest with you, 90 % of the people tagged themselves out and said, I don't have enterprise selling in my DNA.”
Listen to the episode Sales team, hiring & comp Link to this
An SMB deal takes about three months to close, while an enterprise deal takes 12 to 18 months. Listen
Ann says she insisted that the executive team and board understand this difference, raising it even with board members during her interviews, because her pipeline was empty and she needed runway. She says that although many agreed up front, they still had a hard time buying into it along the way and were asking when results would come nine to twelve months in.
“an SMB deal is going to take about three months. An enterprise deal is going to take 12 to 18.”
Demand fulfillment means buyers already have a budgeted project, while demand generation means selling a story to buyers who have no project yet, and the latter adds at least six months to an enterprise deal. Listen
Ann says demand fulfillment involves prospects who come in or are reached and already have budget and a project, closing in about three months. Demand generation involves selling a crafted story of value to people who have neither a project nor budget, which she says adds six months or more to an enterprise cycle.
“demand fulfillment is people are coming into us or we're talking to people and they're like, yes, we've got a budgeted project.”
Listen to the episode Pipeline & demand generation Link to this
Crunchbase measures its funding-round calls on precision and recall rather than accuracy, and that calling 85 to 90% of rounds two to six months ahead at 92% correctness would be valuable to investors. Listen
Ann says the output is currently called predictions and insights but she wants to move away from that word, because financial services firms regard prediction as their own job. She says the measure is how much of the total was found and how much was missed. She frames the 85 to 90% and 92% figures as an 'if' example of the value to someone deciding which markets to invest in.
“if we're calling 85 % or 90 % of the funding rounds, two to six months in advance, and we're right 92%”
Ann suggests bundling enterprise data access with a usage component, such as roughly 40 cents per query, because companies rarely buy enterprise licences priced on headcount. Listen
Ann says companies will not buy enterprise licences for software by reference to headcount, such as 500,000 or 200,000 employees. Her proposed alternative is a bundled price with a usage fee layered on top, for example paying 40 cents a query through an MCP.
“Companies will never buy enterprise licenses to anything normally because I've got 500 ,000 employees or I've got 200 ,000 employees.”
An AI tool will not replace enterprise sellers on $500,000-plus deals, because these require coordinating many stakeholders, though small deals can be fully automated. Listen
Ann says enterprise deals involve a lot of interconnected people at the buying company, and the seller is responsible for getting them all aligned. She says that is unlikely to change at the enterprise level, while a $5,000 deal can be automated.
“When you're talking enterprise, I'm talking $500 ,000 plus type deals.”
Sellers should stay in control of the conversation, because customers who drive the deal can take the seller down unproductive paths. Listen
Ann tells her teams that the seller is the driver of the deal, not the customer, and that letting the buyer steer leads to dead ends. She frames time as the most valuable asset a salesperson has, and says busy does not mean successful unless that time is used diligently.
“You are the driver of this, not the customer. So make sure you're in the driver's seat and you're not just being driven by them because they'll take you through all the windy roads and end up at a dead end.”
AI is only as good as the data it is built on, and public LLMs mostly reflect public-domain information. Listen
Ann argues that public LLMs act like a better search engine for public information, while large AI players are gathering proprietary data to differentiate. She says customers and AI partners are trying to combine their own data with industry-specific proprietary datasets, citing financial services as an example.
“AI is only going to be as good as the data that it's built upon.”
Companies should combine their internal data, such as emails, chat and documents, with the right third-party datasets to get a useful AI system. Listen
Ann says a company's main proprietary data is internal material like emails, Slack and documents, but that AI tools only produce decision-grade output when the right third-party data sources are connected. She says without those sources, business decisions cannot reliably be made from the output.
“you need to take all of that and mix it with the appropriate third party data sets to really bring forth the most robust and successful AI experience.”
Crunchbase's competitive position rests on its private-market dataset plus visibility into how its users interact with the software. Listen
Ann says Crunchbase has more private-market data from inception through Series K than anyone, covering company profiles, competitors, investors and funding rounds. She adds that its 80 million-plus software users give the company behavioural signals that others cannot access, which lets it judge whether a company is gaining or losing momentum.
“having all this data and our 80 million plus users of the software and being able to watch their digital body language that nobody else has access to”
Private markets now make up roughly 75 to 80% of the Nasdaq's market cap, which she cites as a reason private-market data is increasingly important. Listen
Ann gives this figure with a hedge, saying she thinks private markets are up to around 75 or 80% of Nasdaq market cap now. She links it to the growth of very large private funding rounds, which she says she has never seen before.
“I think they're up to like 75 or 80 % of the Nasdaq market cap now.”
Looker's edge came from its data model and semantic layer, and AI natural-language querying now sits on top of that, replacing dashboards built on dashboards. Listen
Ann says Looker's data model and semantic layer governed how data was managed and maintained, and that natural-language querying became the analytical layer on top. She cites Google's acquisition of Looker, at a price of $2.6 billion on a roughly $100 million run rate, as evidence of that value.
“it was about the data model in the semantic layer of how the data is managed and maintained”
She repeats her enterprise sales timeline to the board, because directors who meet quarterly do not understand why enterprise deals take so long. Listen
Ann says board members spend only a couple of hours a quarter on the business, so they often lose track of the reasons behind a long sales cycle. She says she keeps reinforcing the same messaging over time to keep them aligned.
“They have no idea why it takes so long.”
Ann started four revenue streams over the past year and now sees one clear winner, while another needs adjusting. Listen
Ann says she developed about four revenue streams over the past year, some with less friction and less dependence on people than others. She says she started all four and is now seeing one clearly outperform, while another needs adjusting.
“I started all four. And I'm starting to see there's definitely a clear winner in it.”
When presenting a plan to the CFO, Ann framed direct sales as the most expensive channel while showing how other channels contribute revenue at lower cost. Listen
Ann says she explained to the CFO that direct sales is the most expensive channel but would be focused on enterprise accounts, while other revenue streams would contribute a share of revenue at a lower cost. She says partnerships mean not every deal is touched by every channel.
“direct sales is probably our most expensive channel, but they're going to be focused here.”
For a pivot into new work, Ann does not want sellers who rely on a playbook and price list, because she is no longer in the transaction business. Listen
Ann says she wants people who are comfortable with ambiguity, can change things on the fly, and can create new deal structures. She contrasts this with sellers who want a playbook and only complete transactions, and says Crunchbase lets customers buy access to the MCP directly from the website for the transaction-type segment.
“I don't want the people that want like the playbook and the price list and they're just going to do the transaction because we're not in the transaction business anymore.”
Listen to the episode Sales team, hiring & comp Link to this
A seller selling behind a well-known brand gets meetings far more easily than one selling behind an unknown name. Listen
Ann says every financial services firm knows the Crunchbase brand, though many know the old Crunchbase and its value proposition has changed. She sees the brand as secret sauce because it lets the team go in and present the new Crunchbase. She contrasts selling behind a Google name with selling behind an unknown company, which she says makes meetings much harder to get.
“selling behind a Google name to get a meeting is so much easier than selling behind ABC that nobody knows”
Software vendors can license their data within the application to create a second revenue stream as their user counts decline. Listen
Ann says a software company with valuable data can license that data to enterprises for their AI front ends, so revenue from data may rise even as seat-based software revenue falls. She says this is the strategy vendors should be working out now.
“You can have a way to license the data within that software application and have another revenue stream.”
Software vendors that assume AI will not take their revenue are mistaken, and that an MCP interface is becoming the new SaaS interface. Listen
Ann says Crunchbase took a hit to its own revenue from AI, which it had anticipated, and that she would be surprised if any SaaS business was not seeing one. She says the SaaS companies that are fine will be those attached to AI companies in some way.
“We've already got our MCP on the front end and that's the new SaaS.”
Her company's own SaaS revenue took an AI-related hit, and that she would be surprised if other SaaS companies were not seeing one. Listen
Ann says the hit to Crunchbase's revenue was anticipated, and the question for SaaS companies is how quickly they can define a go-forward strategy. She says those that think they are too important to lose users will face problems.
“Anyone in the SaaS software business that's not seeing a hit to their revenue from AI, I would be surprised.”
An AI tool can cut a financial analyst's week of market research to about 30 minutes. Listen
Ann describes a financial analyst whose job is to research new markets and companies, and says the job can be done in about 30 minutes rather than a full week. She presents this as an example of the productivity gains from analysing data through an AI interface.
“They can do their whole job. of a full week as research in probably 30 minutes.”
Customer research should start with a few AI prompts on the prospect's 10-K, then be checked with the buyer, asking whether it is accurate and what is missing. Listen
Ann says the 10-K remains the best resource for salespeople researching public companies, though fewer people read it in full. She says sellers should show up with what they understand about the buyer's priorities and ask the buyer to confirm or correct it, because customers do not want to repeat themselves.
“Showing up with this is what I understand. Is this accurate and what is missing?”
Ann agrees with a host that RevOps means different things at different companies, and says a candidate's résumé is the best guide to what they will actually do. Listen
A host said RevOps roles vary widely: some people come from sales, tech or finance, and some own the CRM, forecasting, recruiting and onboarding. Ann agreed, adding that she would not expect someone to do anything vastly different from what they have done before. She says she hired and let go two RevOps people because they did not do the analysis she needed, and she now uses an AI front end on Salesforce and Gong for most of it.
“show me someone's resume and I'll show you what to expect from them”
Crunchbase customers use its company-level data to set lending terms, similar to how credit bureaus score individuals. Listen
Ann says customers apply different lending percentages or packages to different companies, in the same way lenders use personal credit scores. She notes that startups often have no traditional credit information, so this kind of data fills the gap.
“we've had customers of ours that say, we're going to give different percentages or different packages of lending numbers to different people.”