As artificial intelligence moves beyond general purpose chatbots and into consumer apps, developers and publishers of AI apps face a more immediate version of an old internet problem: how to pay for the product.
Even as token costs reportedly plummet, AI features still cost far more to operate than traditional web and app experiences. In February, Koah announced a $20.5 million Series A and describes itself as “the ad network built for AI”: a marketplace connecting advertisers with AI apps through native, real-time intent matching. The company believes that makes advertising a necessity much earlier in the life of an AI product.
In a conversation with tipsheet yesterday, Co-Founder and CEO Nic Baird discusses the company’s growth, why conversational ads should be treated as a consideration channel rather than a new version of Search, the measurement challenge created when AI acts as a thought partner and how dynamically generated ad formats could support a wider AI app ecosystem.
Topics covered include:
- Koah’s growth and marketplace momentum
- Koah’s ad network model for AI apps and advertisers
- Why consumer AI monetization is arriving earlier
- The opportunity beyond ChatGPT and Claude
- What ChatGPT Ads is teaching the market
- Which advertiser categories are working today
- AI as a consideration and thought partner channel
- Attribution for conversational media
- Why inference costs constrain AI publishers
- Dynamic formats and generated ad copy
- Koah’s supply-side strategy
- What conversational advertising could look like in three to five years
Scroll down for the interview, which has been lightly edited for clarity.
Funding the AI app economy
tipsheet: Since Koah’s last funding announcement in February, what has changed in the business?
Nic Baird: A lot has changed, but it may be useful to start with the problem Koah was built to solve: consumer AI companies did not have a sustainable business model to cover the much more expensive inference costs that come with AI.
At the time, people thought subscriptions would be the answer and that ads would never really arrive in AI. We didn’t think that was right.
Two years ago, we assumed a wave of AI-native companies would replace old, non-AI companies. Some of that has happened. But the more important realization was that the generative interface itself was going to take over, not necessarily AI-native companies alone.
We raised our last round on the thesis that traditional companies would begin adopting these surfaces. Because inference is expensive, monetization would have to arrive much earlier. We felt the market was at an inflection point.
That has largely proved true. Eight months ago, many larger AI companies — like Character[dot]ai — were reluctant to show ads. Now more companies are showing them, even if some are still using traditional display formats. And chat interfaces are reaching mass market across traditional web and app products. That is where much of our supply-side growth is coming from.
On the demand side, ChatGPT Ads has made advertisers much more sophisticated about what this channel is and what its challenges are. A year ago, advertisers often treated AI as the new Search and moved Search campaigns directly into it. That did not necessarily work. They are learning that this is a consideration channel and a middle-of-funnel channel, not simply another bottom-funnel Search product.
Marketplace momentum
tipsheet: You said on LinkedIn a couple of months ago that Koah added more than $1 million of ARR in May. What can you say about momentum now?
Nic Baird: It has been a very good couple of months. I will not share a lot of specific revenue numbers, but we tripled GMV, or the amount spent in the marketplace, from July to August. We are projected to triple it again from August to September.
Advertisers are understanding this market. They are expanding budgets and moving us into core budgets. At the same time, adoption of AI surfaces is taking off across many types of companies. We have lofty projections for the end of the year.
In a follow-up request by tipsheet, Koah said AI app publishers typically receive 70% of advertiser spend. The ad runs where Koah’s system determines it is most likely to convert.
tipsheet: What is driving that momentum?
Nic Baird: Demand has always been relatively accessible for us. But it is easier than before to become an always-on or core part of an advertiser’s budget because advertisers now understand the market. They are testing in ChatGPT and treating it as a core motion.
On the publisher side, consumer adoption is growing. More people are using AI in ways that are commercially relevant. The hallucinations and problems publishers worried about a year ago have also largely diminished, so publishers feel safer rolling these features out to more of their users.
Beyond the largest chatbots
tipsheet: Most users are spending time in ChatGPT and Claude. What does the opportunity look like beyond those largest assistants?
Nic Baird: We have learned that parts of our network reach incremental AI users that ChatGPT and Claude do not. But many people also use multiple tools. That is a meaningful difference from Search: it is very liquid to move between models and products.
The product interface built on top of a model can create incremental value. A strong AI shopping assistant, travel planner or emotionally-focused assistant can complement a person’s use of a general purpose chatbot.
The opportunity is incremental users and incremental use cases. And the chat interface is spreading into traditional products. Companies such as Time and Reuters are beginning to roll out these features. We are seeing the diversity of surfaces accelerate.
The ChatGPT Ads learning curve
tipsheet: What is your assessment of ChatGPT’s advertising evolution?
Nic Baird: People thought it would be easy: match the context of a query, capture the intent and treat it like Search. It is not that simple.
Formats are complicated. Making them native is difficult. Building liquidity in the marketplace is hard. To be relevant across the breadth of ways people use AI, you need a huge diversity of products and SKUs. Measurement is also difficult because this is not always a bottom funnel channel with clean last-click attribution. You have to think about consideration, lift and incrementality.
OpenAI is doing an amazing job. It is moving incredibly fast. The product is still nascent and there is much to develop, but some advertisers and verticals work immediately while others need different formats, targeting or landing page approaches. A carousel could unlock one commerce vertical where an image format does not. That is the work ahead: formats, targeting, relevance, post-click attribution and the presentation of information.
Where conversational ads work today
tipsheet: What are the early sweet spots for advertisers?
Nic Baird: Early AI users tend to skew younger and more tech native, so technology products, software, games, entertainment and devices tend to perform well.
The other useful question is what behavior people exhibit in these spaces. A large share of AI use involves homework help, writing help and understanding problems. There are a lot of students. Another meaningful portion has direct or indirect commercial intent, perhaps 10% to 15%, which may be similar to Search.
The challenge is that a conversation can begin as loose problem-solving and develop commercial intent over time. Products that solve social or interpersonal problems can perform well. We see categories such as weight-loss drugs, personal finance and image generators perform. Social products can work well too.
AI as a thought partner
tipsheet: How does the purchase funnel look different in conversational media?
Nic Baird: The question for us is where AI has the biggest effect on whether someone ultimately makes a purchase. We would love to bring the whole store into the chat interface and complete the transaction there, because then attribution is straightforward. But that is not always how people make decisions.
For most people, AI is more of a thought partner. Google is a great commercial channel because it is navigational. Someone searches for Nike and goes to Nike[dot]com to buy the shoe. AI is closer to a trusted source or advisor.
I have been researching an e-bike for months. Claude helped me understand the specs and trade-offs and identify the store I should visit. I will probably buy the bike there. But how does Claude get credit? That is the hard problem, particularly for higher consideration products with longer journeys. AI is affecting decisions, but measuring it is difficult.
For smaller purchases, agentic commerce could reduce friction over time. Shopify, Amazon and Google are already extremely efficient at moving a person from intent to purchase, including the returns process. AI has no business competing with that today. Eventually, ChatGPT and others may develop stronger commerce ecosystems, enabling more transactions inside the surface and cleaner last-click attribution. For higher consideration products, though, we have to understand AI’s influence further up the funnel.
Inference costs and the publisher opportunity
tipsheet: Why are publishers coming to Koah?
Nic Baird: One of our favorite publishers is an app called Daimon. It had to slow growth for three or four months to optimize inference costs so it could grow without bankrupting the company.
That is representative of the market. Bigger players trying to launch AI surfaces often do not want to take them out of beta because they are too expensive. They may be reluctant to offer the feature to half their users. The concern around hallucinations is diminishing, but the cost is still dramatically higher than hosting a normal website or app.
In the earlier internet and mobile eras, companies could focus on growth and distribution first, then monetize later. Meta did not turn on ads until it had a billion users. ChatGPT did not turn on ads until it had a billion users, but it also had access to enormous amounts of capital. Most publishers do not.
Our mission is to make sure the world is not just three AI companies. We want a publisher to be able to build a surface that is valuable to 50,000 people with a particular use case or affinity and not worry that it cannot scale because it will not cover its costs.
Dynamic formats and generated copy
tipsheet: Koah has positioned itself around native advertising rather than standard IAB formats. How does creative work today?
Nic Baird: From the advertiser perspective, the lens is creative. From the publisher perspective, we think about formats.
We are not trying to create an IAB standard. AI is dynamic, and the format should be dynamic too. It should personalize to the user and take the shape it needs based on the intent of the query. A top-of-funnel experience should look different from a bottom-of-funnel one.
Our formats are largely generative. They can change shape and select different components. We have an engine that uses hundreds of components generated over the life of the company and assembles the format that makes sense for a given query.
On copy, we can run static creative like any other ad platform. But when we generate copy on the fly, we see close to a 2x performance lift. It allows the message to explain, in real time, why a product solves the user’s particular problem.
We are not yet generating images or video ourselves. Advertisers sometimes use generative creative tools. Over time, we may bring that into the product, but today we are focused on text. About 50% to 60% of our ads are text-only, and text works well when the intent is there.
What comes next
tipsheet: Where is conversational advertising today, and what should the market expect over the next three to five years?
Nic Baird: Today, there is a big influx of testing spend as advertisers try to learn the market. It is still immature, even though the progress in a short period has been remarkable.
The big questions are formats, targeting, relevance and attribution. In three to five years, I think those will be much more clearly answered. Advertisers will know where the channel is most impactful and how to capture value after the click. Platforms will know how to target better and show content in the right format. The attribution story will become much clearer, and marketing will become more sophisticated about getting the actual value out of AI rather than trying to force it into the Search mold.
tipsheet: Where is Koah focused now, and what would success look like a year from now?
Nic Baird: Our focus is supply-side growth: expanding the number of people who can build these interfaces, whether they are AI-native apps or publishers putting AI features on their websites. We want to accelerate those surfaces with technology, advertisers and high RPMs.
In a year, we would like to have a strong, data-supported view of the most valuable way advertisers can use this channel. We believe the middle of the funnel and consideration are where it shines, but nobody has fully drawn the line from that hypothesis through to conversion.
For publishers, we want the answer to be simple: when someone thinks an AI interface is too expensive to scale, they think of Koah. We want to prove that personalized, relevant advertising can fund growth without harming the user experience. Some of our data suggests it can improve both seven-day and per-session retention. We want to build on that and help publishers create the best possible experience.