AI futures no 1 - Semantic-first websites and the 'feed the model' Org OS

We will be designing websites ‘AI first’ pretty soon and this will be as common as designing for SEO now. Arguably it will outgrow SEO as SEO itself will be subsumed by AI discovery optimisation as a practice and strategic need. Of course designing for semantics is nothing new but it will become the primary way we think about the provision of digital information.

Organisations who provide advice and/or information as a service will realise that their encapsulated knowledge is their service (literally, their service model) -- and also that it is the easiest thing to rip off. Organisations will have to make decisions about whether providing content for AI is part of their service or not. The answer will likely be ‘yes’ (or risk irrelevance) and this will lead digital provision to be semantic first. Organisations will have to lean into the data world and will have to create machine-learning approaches to knowledge management, most likely in the form of custom models. This also has implications in knowledge capture. You would want to have ‘graph everywhere’ style knowledge models.

The idea that knowledge/advice will take a year or two -- or even a month or two -- to make it to a website will seem antiquated. It is antiquated already, but the OS of many organisations to minimise risk and allow for quality checks still reigns. This ‘reduction of the middle’ will mean that there will be an increased emphasis on actual conversations and experiences being rapidly digitised in close to real time.

The technical implication of this is that real time audio capture into text will be required to stay relevant, and ‘feeding the model’ will become one of the core requirements of an organisation’s workforce.

Consequently ‘The website’ will be a delivery mechanism for dynamic requests. These structures may look like topic-based websites but the idea that there will be fixed categories and set navigation is quaint. Categories and media rendering will be emergent and contingent. The same core engine will deliver in multiple media and domains. That core engine will be the organisation. It’s also worth noting that the core engine will not be ‘structure + data + rules’. It will potentially be opaque and those rational data schemes won’t work, another organisational challenge in the risk domain.

All of this will have an effect on who organisations will employ. Anyone on the ground talking to clients will be more important than head-office in-house subject experts, of whom we will be needing a lot less. In house experts will be made irrelevant by in-house expert systems. This will reverse the centralising tendency we have seen where ‘brand’ and control is more important than actually helping people.

The key role for websites (and other repositories of organisational knowledge like handbooks, guides, magazines, annual reports and so on) will be as the AI starter, the basis for the first model. Then the model will take over and ‘the website’ will be a functional ruleset on one ‘pipe’ from the model. Initially that pipe will be formally similar to what we have now (essentially a ‘smart’ headless model) but that too will evolve into a pipe that can be provisioned and run by a domain specific publishing AI.

As an aside this is what ‘Digital transformation’ will look like in a few years - if you haven’t led that or a similar project then you won’t really be able to label yourself ‘Digital change agent’ anymore. Just saying.