AI’s tricky path for content & software providers

The most powerful AI models accessible to everyone are Large Language Models offered by OpenAI, Anthropic, Meta, Google, and others. Besides very talented engineers and significant hardware resources, they all need the same thing; data. This is not only important for making the models better. It can also provide the context that is unique to your organization. Having access to internal company data opens up a whole new world of AI use cases.

It’s no surprise that AI vendors are moving quickly to get access to more data. For example, OpenAI offers news publishers between $1M-$5M per year to use their content to train models. These are the same publishers that successfully made the crucial transition from print to digital and later again from web to mobile. Should they sign off their unique content to the next defining wave of digital, with a brand new shape of user interaction in conversational AI? 

AI players also want access to company data

The above example is mostly about already publicly available data, with a portion of paywall-protected content. But the need for data does not stop here. It extends to getting access to internal company data. Microsoft’s Co-Pilot now integrates with Salesforce and Zendesk, to access your company’s customer, sales, and support data. The same is happening for development teams through Jira. Software vendors can create plugins and get company data into Co-pilot, so expect this to happen for many of the SaaS tools your company is using. 

Why this is risky for content & software players

The above examples all seem fine because it’s about increasing your internal team’s performance. However, if you are a content or software provider, allowing your users to use an external AI conversation interface can bring significant risks to your business:

• Instead of owning all direct interaction with your customers from your products & channels, suddenly a chat interface from another company sits in between. 
  ‍
• New competitors can emerge. An AI vendor now can start building workflow and other smart features around data and that normally only was available through your offering.
  ‍
• It is very hard to control where your data might go. Once it's part of an AI conversation, it can also be accessed & used by other players that have integrations on the specific AI platform.

Integrate versus build-it-yourself

Should content or software providers integrate with Microsoft Copilot, OpenAI ChatGPT and others? We think this should be considered very carefully if this fit's your strategy. The upsides of visibility and distribution might not outweigh the possible downsides of losing a portion of direct customer interaction, new emerging competitors and giving up control over your unique data.

We have worked for years in the app store ecosystem and have seen the constant friction that Apple created by putting itself between the app company and the customer for billing. Ultimately resulting in lawsuits, which was only about the billing relationship. Integrating with the large AI companies' chat products moves the whole user experience to another player, which has a much bigger impact.

One of the solutions is to build-it-yourself. AI technology has never been more accessible. There are ample tools and models you can use to build a great AI experience in your existing product, for your unique content & data. Yes, it requires a new technology stack and engineering practices. But every new big technology wave has required this. It was the same for the app ecosystem, and we know how that has proven to be a very successful path for companies to invest in.

The build-it-yourself versus integrate is already happening. The three players mentioned above that first integrated with Microsoft Co-pilot, now all also have their own conversation AI offer, built into the Salesforce, Zendesk, and Atlassian products.

It’s all about the customer workflow

The company that owns the interface between your unique data and the customers has the opportunity to build what any content or software vendor ultimately wants; move into the customer workflow. Not just be a tool that is used now and then, but become indispensable in daily work. AI is the biggest opportunity and a threat at the same time for this. 

We think there are clear steps on how to capture that opportunity, instead of the downside:

1. Build AI features in your own product, on your own content and data.
   ‍
2. Do not integrate with AI vendors to bring your content and data to their user-facing platform.
   ‍
3. Be rigid about opening up your data outside your platform, and start introducing very clear API terms, to prevent other players from building AI vendor integrations using your API.

These actions are very much possible with tools out there today. Your unique content & data is the biggest asset for the future that AI is bringing. Move fast, but keep your internal data very, very close.

‍