The Old Knowledge Business Is Dying – Expertise Isn't

01 /The Concept

Scaling yourself has always been the impossible constraint on expertise. Delphi thinks AI has finally made it possible.

To train it, you feed it the content you've already created – in any format: messages, forum replies, blog posts, newsletter archives, Notion notes, audio recordings, webinar recordings. The more, the better.

Among Delphi's early high-profile clients is Lenny Rachitsky, the well-known product newsletter writer. Arnold Schwarzenegger recently created his own twin on the platform, focused on helping people build discipline – appropriate, given that his physical accomplishments are a direct product of it.

Delphi's target audience is coaches, creators, consultants, and public figures who sell or share their expertise, perspective, and worldview with others.

The core problem is that personal communication doesn't scale. Your time is finite. A digital twin can engage with any number of people, around the clock.

The key functional advantage: the twin answers specific questions. Everything it knows is technically discoverable in an expert's published content – but hunting through years of writing or recordings is its own time investment.

The result is that users get exactly what they want without consuming the expert's time. Experts can sell that access as a standalone product or offer it as an added benefit to existing subscribers.

A Delphi twin can be connected to any communication channel – a website, Slack, WhatsApp, or Zoom – and converses in any language, regardless of what language the original content was in.

Free accounts are available but with limited source content and text-only chat. At $99/month, users can upload up to 5 million words and enable voice conversation. At $499/month, the limit rises to 12 million words with video capability added. At $2,499/month, source content is unlimited and the twin runs on an advanced reasoning architecture.

Delphi opened its beta in December 2023, raising $2.7M for that launch. It has now closed a new $16M round led by Sequoia Capital.

02 /Why It Matters

Delphi frames its platform as a new medium for transmitting knowledge – positioning itself alongside the printing press, radio, television, and podcasts.

The analogy holds in a structural sense. But the current positioning – as a tool for experts, creators, and celebrities to scale their personal brands – raises questions. Several startups that raised money on almost identical premises have since pivoted away from it.

Personal AI – [covered here](/review/vzorvat-rynok-obrazovanija) – launched in 2023 with a platform for building a "digital version of yourself" and raised $13.7M on that premise. Today it sells the same platform to companies building digital twins of their employees, to "multiply" internal teams. Two subsequent funding rounds were raised after that pivot, both undisclosed.

Amigo – [covered here](/review/ii-experty-eto-sovsem-ne-ii-sotrudniki) – raised $6.3M in November 2024 on a Delphi-like platform for experts and consultants wanting to scale. It now sells to medical clinics that want digital twins of their doctors, nurses, and patient-facing staff.

Why does this keep happening? The answer is that experts and consultants don't actually sell information – they sell a personal brand. Clients aren't paying for the advice in the abstract; they're paying for advice that comes from a specific person.

A fashion analogy: a Birkin bag starts at $20,000. High-quality replicas sell for $300–500. But Birkin buyers don't buy replicas – not because the quality differs meaningfully, but because they're paying for the brand, not the object.

Same logic applies here. A person who follows a coach charging $1,000 an hour isn't going to pay $10 an hour for a digital replica of that coach, even if the answers are substantively identical. They're paying for access to the person, not the information. If the price is too high, they'll find a less expensive expert with a personal brand they trust – not a replica of someone they can't afford. Just as people who can't justify Birkin pricing buy bags from more accessible brands, not convincing fakes.

Put simply: people appear to perceive digital twins of public figures as counterfeits – even ones created by the originals themselves – and that perception suppresses both demand and willingness to pay.

Which raises the obvious question: why would any expert create a digital twin that might dilute or damage their brand? It would be like Hermès selling both original Birkins and licensed replicas simultaneously. The likely outcome: buyers lose confidence in both.

03 /Opportunities

The tension here is genuinely interesting. On one side, influencers and experts are advised to build strong personal brands. On the other, a strong personal brand actually limits scalability – because people only pay for direct access, and direct access is inherently finite.

That leaves two paths, each of which points toward a distinct startup opportunity.

The first: corporate platforms for digital twins of employees, enabling companies to scale internal or external communications without adding headcount. That's the path Personal AI and Amigo landed on after pivoting.

The second: stop scaling "blah blah blah" and start converting expertise into products that produce measurable outcomes. Stop selling the brand, start selling the result. A few examples.

One person spent years in consulting and course sales, helping companies write better proposals and tender submissions. When AI arrived, he channeled that expertise into AutogenAI – [covered here](/review/prostoj-sposob-ubedit) – a platform for generating compelling bids and proposals. The startup raised $65.3M across two rounds in the six months after launch.

Another person worked as a CMO and built an internal AI platform for search optimization that delivered strong results. He began running paid workshops teaching others to replicate it. Attendees were impressed – but nobody actually implemented it on their own, because setup and ongoing management would have been too much of a hassle. So he left the company, built his own startup GrowthX – [covered here](/review/novaja-biznes-model-dlja-bystrogo-i-pribylnogo-rosta) – and started delivering SEO-as-a-service using the same platform with his own team running it. Revenue grew from $0 to $600,000/month between July last year and April this year. In May, GrowthX raised $12M in its first funding round.

The opportunity for experts: stop selling the expertise itself and start building products that execute it – products that return something measurable to users.

The opportunity for product builders: find experts with genuine, codifiable knowledge and help them make that transition. Not all experts have the skills to build products from what they know.

If the right category of expert can be identified, there may even be a case for a platform that converts certain types of expertise into products automatically – or at least semi-automatically.

This second path is interesting precisely because it scales. There are enormous numbers of genuinely skilled people out there. And among them, there are certainly those whose expertise can be structured, codified, and turned into something that actually works for other people.

In other words, you could build something conceptually adjacent to today's Delphi – but with a fundamental difference: instead of scaling conversation, it scales outcomes.

What types of expertise in which domains are most amenable to AI-driven codification? What products could emerge from each? What results could they deliver, and what would they need to look and do to deliver them?

At the corporate level, some startups are already pursuing a version of this – turning traditional consulting into product businesses. Instead of decks full of recommendations, they deploy AI products that actually implement those recommendations and produce real results for clients. One such startup called this model "Consulting 2.0." Examples include Gruve – [covered previously](/review/kak-sovmestit-pribylnost-i-masshtabiruemost), which raised $37.5M – Quantum Rise, and Operand, which graduated from Y Combinator this spring and raised $3.1M in May.