Latest / Uphoff on Media Podcast / If Your Moat Was Data, What Happens When AI Can Manufacture It?

Transcript

If Your Moat Was Data, What Happens When AI Can Manufacture It? · Uphoff on Media Podcast

• 0:02
  SPEAKER_00: Welcome to Uphoff on Media.
• 0:04
  SPEAKER_00: I'm Tony Uphoff.
• 0:05
  SPEAKER_00: In late 2021, Zoom info was worth about $25 billion.
• 0:10
  SPEAKER_00: Today, it's worth $1 billion.
• 0:13
  SPEAKER_00: The database didn't get worse.
• 0:14
  SPEAKER_00: The customers didn't vanish.
• 0:16
  SPEAKER_00: So what the heck broke?
• 0:18
  SPEAKER_00: My argument is that the data moats have been built in three eras.
• 0:22
  SPEAKER_00: In the first, the moat was the cost of assembling the data.
• 0:26
  SPEAKER_00: In the second, it was scarce data living inside software customers couldn't easily leave.
• 0:31
  SPEAKER_00: The third, the agentic era, removes the foundation of both.
• 0:36
  SPEAKER_00: Agents can now assemble facts for close to nothing
• 0:39
  SPEAKER_00: and customers can use those same agents to build the workflow themselves.
• 0:43
  SPEAKER_00: Synthetic data adds pressure on anyone selling patterns instead of facts.
• 0:50
  SPEAKER_00: I ran thomasnet.com through a version of this and it came down to two questions.
• 0:54
  SPEAKER_00: Can a customer get your data anywhere else?
• 0:57
  SPEAKER_00: And if they can,
• 0:58
  SPEAKER_00: what did you build on top of it that they'd keep paying for even if
• 1:02
  SPEAKER_00: the data were free?
• 1:05
  SPEAKER_00: So what's the answer in the agentic era?
• 1:07
  SPEAKER_00: Stop selling access, start delivering results.
• 1:11
  SPEAKER_00: That changes your services strategy, your channel strategy, and how you architect your product.
• 1:17
  SPEAKER_00: The post has two charts to help.
• 1:19
  SPEAKER_00: One maps the three errors.
• 1:20
  SPEAKER_00: The other is the data mode test,
• 1:22
  SPEAKER_00: a two-by-two that shows you which square your business is actually in.
• 1:27
  SPEAKER_00: And there are three questions you can answer in a sentence each.
• 1:30
  SPEAKER_00: Run them on your own business before an investor does it for you.
• 1:34
  SPEAKER_00: Okay, let's get to work.
• 1:35
  SPEAKER_00: This episode is titled, If Your Moat Was Data, What Happens When AI Can Manufacture It?
• 1:45
  SPEAKER_00: In late 2021, Zoom Info was roughly $25 billion.
• 1:49
  SPEAKER_00: Today, it's worth about a billion.
• 1:51
  SPEAKER_00: The database didn't get worse.
• 1:53
  SPEAKER_00: The customer list didn't disappear.
• 1:55
  SPEAKER_00: Something else broke,
• 1:56
  SPEAKER_00: and every B2B data and intelligence business needs to understand what before
• 2:01
  SPEAKER_00: the market explains it to them the hard way.
• 2:05
  SPEAKER_00: The company says the moat still holds.
• 2:07
  SPEAKER_00: ZoomInfo cut its full year revenue guidance in May, then raised it modestly in August.
• 2:12
  SPEAKER_00: Net revenue retention has slipped from 90% to 89%.
• 2:16
  SPEAKER_00: Existing customers are generating 11% less revenue than they were a year ago.
• 2:21
  SPEAKER_00: The stock has fallen from a peak, a 2021 peak of $77.35 to under $4.
• 2:30
  SPEAKER_00: CEO Henry Schuck's response has been direct.
• 2:33
  SPEAKER_00: In May, his quote was,
• 2:35
  SPEAKER_00: Zoom info data and our go-to-market context is the ultimate competitive advantage.
• 2:40
  SPEAKER_00: In August, he stated, you cannot token your way to building the data asset that we have built.
• 2:47
  SPEAKER_00: The first is an argument about where the value sits, data plus context, not data alone.
• 2:52
  SPEAKER_00: The second is a scarcity claim, and it's the one most exposed to what's coming.
• 2:57
  SPEAKER_00: Whether either holds is the only question that matters here,
• 3:00
  SPEAKER_00: and it's not a question the company gets to answer for itself.
• 3:03
  SPEAKER_00: The market answers it.
• 3:06
  SPEAKER_00: I've had to answer these questions from the inside, not from a research note.
• 3:11
  SPEAKER_00: I spent five years leading thomasnut.com,
• 3:13
  SPEAKER_00: an industrial data and directory business,
• 3:15
  SPEAKER_00: through the transformation that turned it into
• 3:18
  SPEAKER_00: a platform Xometry paid roughly $300 million to acquire.
• 3:23
  SPEAKER_00: Every year, that transformation came down to two questions about our own assets.
• 3:27
  SPEAKER_00: The first was scarcity.
• 3:29
  SPEAKER_00: Could a customer get this data anywhere else or was it just expensive
• 3:33
  SPEAKER_00: for someone else to assemble at the time?
• 3:35
  SPEAKER_00: If it was only expensive, we had a cost advantage and cost advantages can erode.
• 3:42
  SPEAKER_00: The second was functionality.
• 3:43
  SPEAKER_00: Had we built enough into the workflow tools our customers ran their jobs through
• 3:48
  SPEAKER_00: that they would keep paying even if the data were free?
• 3:51
  SPEAKER_00: If so, we had a product, not just a data set.
• 3:54
  SPEAKER_00: The answers describe very different businesses.
• 3:57
  SPEAKER_00: The pitch deck reads the same for all of them.
• 4:01
  SPEAKER_00: Data moats were built in three eras, and the third removes what the first two stood on.
• 4:06
  SPEAKER_00: The first era was aggregation.
• 4:07
  SPEAKER_00: The value was in assembling what was scattered,
• 4:10
  SPEAKER_00: a directory of industrial suppliers,
• 4:12
  SPEAKER_00: a database of companies and contacts, a registry of transactions.
• 4:17
  SPEAKER_00: It took people time and capital to build, and that cost was the moat.
• 4:21
  SPEAKER_00: Customers paid because doing it themselves was not a serious option.
• 4:24
  SPEAKER_00: ThomasNet began here as a directory.
• 4:28
  SPEAKER_00: The second era was workflow.
• 4:29
  SPEAKER_00: The data was still scarce or expensive to mine,
• 4:32
  SPEAKER_00: but now it lived inside software the customer used to do a job.
• 4:36
  SPEAKER_00: The sales rep's prospecting screen, the buyer's supplier search.
• 4:40
  SPEAKER_00: Switching meant retraining a team and rebuilding a process.
• 4:44
  SPEAKER_00: Finding another source was the easy part.
• 4:47
  SPEAKER_00: That is the era Thomas Nett's rebuild into a platform belonged to and
• 4:50
  SPEAKER_00: the era ZoomInfo was built for.
• 4:54
  SPEAKER_00: The third era is the agentic era, and it attacks both foundations at once.
• 4:59
  SPEAKER_00: Agents can now assemble facts from public sources for close to nothing,
• 5:03
  SPEAKER_00: which erodes the cost of aggregation.
• 5:05
  SPEAKER_00: Customers can use those same agents to build the workflow layer themselves,
• 5:09
  SPEAKER_00: which erodes the switching cost of the second era.
• 5:13
  SPEAKER_00: Shuck says it has started.
• 5:15
  SPEAKER_00: Customers are building interfaces that combine their own data with Zoom Infos,
• 5:19
  SPEAKER_00: though so far only at, quote, unquote, the top less than 1% of companies.
• 5:25
  SPEAKER_00: That is where a pattern shows up first.
• 5:29
  SPEAKER_00: Agents assemble data and models generate it and they threaten different businesses.
• 5:34
  SPEAKER_00: AI reaches a data business by two different routes and
• 5:37
  SPEAKER_00: the second gets less attention than it should.
• 5:40
  SPEAKER_00: The first is agentic assembly.
• 5:42
  SPEAKER_00: An agent researching a prospect can crawl the open web,
• 5:46
  SPEAKER_00: pull public records, and structure what it finds in seconds.
• 5:49
  SPEAKER_00: Clay lets a customer run a single lookup across many data providers in sequence,
• 5:54
  SPEAKER_00: which turns a database into one input among several.
• 5:58
  SPEAKER_00: Nothing gets fabricated.
• 5:59
  SPEAKER_00: Real facts are collected cheaply.
• 6:01
  SPEAKER_00: If the capital needed to assemble and refresh a contact database was the real source
• 6:05
  SPEAKER_00: of the moat, then the moat was a temporary cost advantage all along.
• 6:09
  SPEAKER_00: And AI removing that cost exposes what the moat was made of.
• 6:13
  SPEAKER_00: This is the most direct pressure on contact and firmographic data.
• 6:18
  SPEAKER_00: The second is synthetic data.
• 6:21
  SPEAKER_00: Records generated by a model with no real-world source at all,
• 6:24
  SPEAKER_00: it's gaining steam in training,
• 6:25
  SPEAKER_00: testing,
• 6:26
  SPEAKER_00: and modeling,
• 6:26
  SPEAKER_00: and businesses that sell that kind of data feel it first because
• 6:30
  SPEAKER_00: a statistical pattern is exactly what a model can reproduce.
• 6:34
  SPEAKER_00: What a model can't reproduce is a fact about a specific real entity,
• 6:39
  SPEAKER_00: who owns an asset,
• 6:40
  SPEAKER_00: what a particular supplier can actually make,
• 6:43
  SPEAKER_00: what a transaction cleared at, and it needs real data to stay anchored.
• 6:47
  SPEAKER_00: A 2024 paper in the journal
• 6:52
  SPEAKER_00: Nature found that indiscriminate training
• 6:55
  SPEAKER_00: on model-generated content causes irreversible defects in the resulting models in
• 7:00
  SPEAKER_00: which tales of the original content distribution disappear.
• 7:04
  SPEAKER_00: The rare cases go first, and rare cases are where a lot of B2B data earns its price.
• 7:11
  SPEAKER_00: Real-world facts can be assembled cheaply but not fabricated so their value moves
• 7:16
  SPEAKER_00: to whoever can verify them and stand behind them.
• 7:19
  SPEAKER_00: I made the upside of that argument in the post The Vertical Intelligence Company.
• 7:26
  SPEAKER_00: Every data business rests on scarcity or functionality, and the agentic era tests both.
• 7:31
  SPEAKER_00: Scarcity and functionality are the whole test.
• 7:35
  SPEAKER_00: In the blog post accompanying this podcast on Uphoff on Media on Substack,
• 7:41
  SPEAKER_00: there is a downloadable framework that you can use as a diagnostic tool
• 7:45
  SPEAKER_00: in your business.
• 7:48
  SPEAKER_00: Both can be true at once, and the strongest data businesses run on both.
• 7:52
  SPEAKER_00: Most don't.
• 7:53
  SPEAKER_00: Most run on scarcity that was really just cost dressed up as exclusivity for as long
• 7:58
  SPEAKER_00: as cost stayed high.
• 8:00
  SPEAKER_00: That's the quadrant Zoom InfoZone numbers suggest the market has placed it in.
• 8:04
  SPEAKER_00: Borrowed time,
• 8:05
  SPEAKER_00: contact and firmographic data that steadily gotten easier for a horizontal AI tool
• 8:10
  SPEAKER_00: to approximate, sold to customers who are already spending less for it.
• 8:14
  SPEAKER_00: Shuck's answer is a bet that the business can move to the functionality side that,
• 8:19
  SPEAKER_00: quote,
• 8:20
  SPEAKER_00: go to market context,
• 8:21
  SPEAKER_00: end quote, is the workflow layer a customer can't build for themselves.
• 8:25
  SPEAKER_00: It's a real bet.
• 8:26
  SPEAKER_00: It hasn't been proven yet.
• 8:27
  SPEAKER_00: And 89% net revenue retention is the market's current vote on whether it will be.
• 8:33
  SPEAKER_00: In the agentic era, the moat is the work you deliver.
• 8:36
  SPEAKER_00: If the data and workflow can both be built cheaply, what's left to build a moat on?
• 8:41
  SPEAKER_00: The answer is the shift I described in the post from SaaS to services software.
• 8:46
  SPEAKER_00: Moats in the agentic era will be built by delivering a service to customers
• 8:50
  SPEAKER_00: through your platform or product that has a material business benefit.
• 8:54
  SPEAKER_00: Data and software become inputs.
• 8:56
  SPEAKER_00: The result is what the customer pays for, and someone has to deliver it and stand behind it.
• 9:02
  SPEAKER_00: I built a business this way.
• 9:03
  SPEAKER_00: At Pipeline 360,
• 9:05
  SPEAKER_00: we created the demand as a service strategy,
• 9:07
  SPEAKER_00: a service delivered to customers rather than software they had to license or run.
• 9:12
  SPEAKER_00: This model helped drive EBITDA from 25% at inception to nearly 40%.
• 9:17
  SPEAKER_00: Moving from selling access to delivering results changes the go-to-market plan
• 9:21
  SPEAKER_00: in three places, and every vendor with a data asset needs to rethink all three.
• 9:26
  SPEAKER_00: The first, services strategy.
• 9:28
  SPEAKER_00: Services stop being the cost center that supports the software and become part of the product.
• 9:33
  SPEAKER_00: Foundation capitals Ashu Garg and Jayagupta put it this way after a year
• 9:38
  SPEAKER_00: of watching AI-native vendors,
• 9:40
  SPEAKER_00: quote, in enterprise AI, integration is not a post-sale activity.
• 9:43
  SPEAKER_00: It is the product surface, end quote.
• 9:46
  SPEAKER_00: Customers now expect to see the outcome on their own data before they sign.
• 9:50
  SPEAKER_00: That raises the cost of sale and vendors have to plan for it rather than fight it.
• 9:55
  SPEAKER_00: Channel Strategy.
• 9:56
  SPEAKER_00: The channel now includes the platforms where the customer's work happens and
• 10:00
  SPEAKER_00: the partners who can carry delivery along with the license.
• 10:04
  SPEAKER_00: ZoomInfo is building for that reality.
• 10:06
  SPEAKER_00: Its gtm.ai layer is designed to plug into platforms like Salesforce,
• 10:11
  SPEAKER_00: AgentForce, and HubSpot Breeze.
• 10:13
  SPEAKER_00: So the data shows up where the customer's agents already work.
• 10:17
  SPEAKER_00: A vendor that insists the customer come to its interface is competing against one
• 10:21
  SPEAKER_00: that shows up inside the customer's agent.
• 10:23
  SPEAKER_00: And Product Architecture.
• 10:26
  SPEAKER_00: Design the product so customers can build and manage part of the functionality themselves.
• 10:30
  SPEAKER_00: That feels backwards to any vendor who spent a decade building lock-in.
• 10:34
  SPEAKER_00: It's the only way to keep the customers who are going to build anyway.
• 10:38
  SPEAKER_00: Open the APIs,
• 10:39
  SPEAKER_00: expose the rules and data as building blocks,
• 10:42
  SPEAKER_00: and let their agents run on your data and your accountability.
• 10:45
  SPEAKER_00: The customer who assembles an interface on top of your product is still your customer.
• 10:49
  SPEAKER_00: The customer who assembles it on top of someone else's is not.
• 10:54
  SPEAKER_00: Run your own business through the test before an investor does it for you.
• 10:58
  SPEAKER_00: Every founder pitching proprietary data and every investor who's ever funded
• 11:02
  SPEAKER_00: that pitch should be able to answer three questions in one sentence each
• 11:06
  SPEAKER_00: without looking at the pitch deck.
• 11:08
  SPEAKER_00: Number one, scarcity.
• 11:10
  SPEAKER_00: Can a customer get this data anywhere else right now?
• 11:13
  SPEAKER_00: If not, why not?
• 11:14
  SPEAKER_00: A real answer is structural, a relationship, a regulatory position, or a physical presence.
• 11:20
  SPEAKER_00: Number two, functionality.
• 11:21
  SPEAKER_00: If a customer could get the raw data elsewhere,
• 11:24
  SPEAKER_00: what have you built on top of it that they would miss enough to keep paying for?
• 11:29
  SPEAKER_00: And number three, result.
• 11:31
  SPEAKER_00: What measurable business outcome does a customer get from you and would you stand
• 11:36
  SPEAKER_00: behind it in the contract?
• 11:39
  SPEAKER_00: A strong answer to the first two puts you in the real moat.
• 11:42
  SPEAKER_00: The third answer is what carries that position or any other into the agentic era.
• 11:49
  SPEAKER_00: The moat was always a question.
• 11:51
  SPEAKER_00: Zoom Info hasn't lost.
• 11:52
  SPEAKER_00: Not yet.
• 11:53
  SPEAKER_00: Henry Shuck is making a real argument about where the value
• 11:56
  SPEAKER_00: in his business actually sits, and it may turn out to be right.
• 12:00
  SPEAKER_00: But the lesson here isn't about Zoom Info.
• 12:02
  SPEAKER_00: If you run a data business,
• 12:04
  SPEAKER_00: write down your three answers before your next board meeting,
• 12:07
  SPEAKER_00: because the market will rewrite them for you.
• 12:11
  SPEAKER_00: If you don't.
• 12:12
  SPEAKER_00: If you can't answer the first two, what you have is inventory.
• 12:16
  SPEAKER_00: But if you can't answer the third,
• 12:18
  SPEAKER_00: you're selling access to a customer who is learning to do without it.
• 12:23
  SPEAKER_00: Data built the moats of the first two eras.
• 12:25
  SPEAKER_00: The work you deliver will build the moat of this one.
• 12:33
  SPEAKER_00: Hey, that's it for this episode.
• 12:34
  SPEAKER_00: If you found this useful, please do one thing.
• 12:36
  SPEAKER_00: Follow the Uphoff on Media podcast on Apple or Spotify and leave a rating.
• 12:41
  SPEAKER_00: It's the single best way to help another operator find this.
• 12:45
  SPEAKER_00: I'm Tony Uphoff, back at it next week.