Within a few days, the two best-known providers of AI assistants have expanded their products in the same direction: away from the pure chat window, towards a platform on which third-party content, data and applications are integrated.OpenAI is primarily addressing customer contact, Anthropic teamwork.For companies that operate websites, shops or internal systems, this raises the same fundamental question: How must content and data be prepared so that they even appear in these environments and are reproduced correctly?

ChatGPT opens up for company profiles

OpenAI is opening ChatGPT as a platform for companies.Via a new Business Hub, companies can claim their profile, enter product data and see how users interact with them.Structurally, this is reminiscent of industry directories and company profiles in search engines, but it shifts the point of contact into the dialogue with the assistant.

This is supplemented by so-called Business Agents.This enables direct customer conversations within ChatGPT, i.e. interactions that no longer necessarily take place on the company's own website.The launch is initially for e-commerce providers in the USA.Which functions will be available in other markets and when cannot be derived from the information available.

Claude grows into a work platform

Anthropic is pursuing a different focus.Claude is getting Sheets, i.e. spreadsheets, as well as whiteboards and so-called Interactive Connectors.Via these connectors, partners such as Adobe and Atlassian can embed their applications directly in Claude.Instead of switching between tools, teams are to use their familiar apps within the assistant's interface.According to the report, an update for meeting minutes has been announced, but no specific date is given.

The direction is therefore clear: Claude is being expanded into the central work platform for teams, increasingly absorbing the functions of other applications.Where OpenAI seeks the path outward to the customer, Anthropic aims inward at everyday work.

What both developments have in common

Despite different target groups, the pattern is identical.The assistant becomes the interface through which content, data and functions from other providers can be reached.This changes the role of your own website and your own specialist applications: they remain the source, but are no longer automatically the place where users arrive.

Three consequences can already be named:

  • Data quality becomes visible. When product data is entered in a Business Hub or evaluated by an assistant, gaps, outdated prices and inconsistent designations immediately stand out.
  • Interfaces gain in importance. Embedded partner apps and agents require that systems can communicate outward in a machine-readable way, not only via an interface designed for humans.
  • The point of contact shifts. If part of the consultation takes place in the chat window, there needs to be a concept of what information may be given there and where handover to humans occurs.

Preparation without flying blind

Neither announcement forces companies to act immediately.The Business Agents are launching regionally and limited to certain industries, and the Claude extensions initially concern collaboration within the team.It therefore makes sense to prepare in a way that brings benefits regardless of which platform prevails.

In practice, this above all includes homework on your own data.Product information, availability, service descriptions and editorial content should come from a maintained source, ideally from the content management system or a connected product information system, and be delivered from there in a structured way.A TYPO3 system, for example, can not only output content as a website but also provide the same data via defined interfaces.Anyone who separates this cleanly serves new channels without duplicate maintenance.

Equally important is clarifying which information belongs outward.An assistant that conducts customer conversations draws on what is made available to it.Internal calculations, interim states or personal data have no place there.A deliberate decision about which data areas are released is part of the technical implementation and not a downstream formality.

Open points

Several questions remain unanswered based on the reports available.These include availability outside the named launch markets, the specific requirements for data formats and the question of how companies can check and correct the presentation of their content in the dialogue.There is also no information on costs or contract models.You should keep an eye on these points before committing budget.

However, it is recognisable that both providers are taking the same path: the assistant becomes the framework into which other systems have an effect.The fact that partners such as Adobe and Atlassian are following this path suggests that a permanent integration layer is forming here and not just an individual function.

Assessment for web and software projects

For ongoing projects, this above all means: content and product data should in future be consistently separated from their presentation and retrievable via stable interfaces so that, in addition to the website, they can also serve other channels.Anyone who invests today in data quality, clean structures and documented APIs can connect new access channels later with manageable effort instead of maintaining content multiple times.The decisive foundation remains substantive control over one's own data, regardless of which assistant ultimately prevails.

Sources