Own the intelligence, software, and data you depend on

People and companies are building more of their work around AI and software they do not control. A provider can change the model, price, rules, or access. Generic systems also push different people and companies toward the same way of working.

I use “sovereign AI” to mean something simple: the person or company relying on the intelligence should control it. The same principle applies to software and data. They should adapt to their owner, preserve what makes that owner different, and remain portable when a provider changes.

That thesis leads to several projects, not one generic consultancy. Each project has a different customer job and must get better at a different repeatable process. I am actively building Aether Many and CRM From Within. Works Like Us and Company Native are earlier concepts.

Aether Many is the agent runtime. It is meant to let each person or company run agents that learn, build abilities, and cooperate while remaining accountable to their owner.

CRM From Within builds an SME's first CRM around how that company already works. It can also build the better-fitting replacement before helping a company move out of an existing CRM. The company owns the resulting software.

Works Like Us explores company intelligence. It would fine-tune open models on approved examples of one company's work, run the model on infrastructure the company controls, and improve it through feedback.

Company Native explores software migration beyond CRM. It would replace existing generic SaaS with custom software while carrying over data, workflows, and users. It stays separate because migration risk and retraining are different work from building a CRM or training a model.

The common thread is ownership. The reason for separate brands is specialization: agent infrastructure, CRM building, model training, and system migration create different skills that should improve with each user or customer.

I wrote earlier about the path from language models to more general intelligence and why privacy in an agent is a judgment problem.

Get in touch

• If you want to build, fund, or use one of these ideas, tell Yusuf's assistant what you have in mind.