In March 2023, Character.AI raised $150 million at a valuation of roughly $1 billion. On October 12, 2026, it filed for Chapter 11 in Delaware with Disney as the stalking-horse bidder, offering $20 million for a non-exclusive perpetual license to its intellectual property and some company claims — a bid that, per the filing, is open to 'higher or better offers.' The platform had millions of daily users. It had a Google licensing deal. It had one of the most-downloaded AI apps in history. What it did not have was a job: a specific, measurable outcome someone would pay to have done. If you are evaluating AI tools for your team this quarter, that is the only line in the story that matters.
1. Engagement Is Not a Business Model — and Neither Is 'AI'
Character.AI was extraordinary at holding attention, and attention turned out not to be the thing enterprises or consumers would fund at scale. The same failure mode is quietly running inside companies right now, with a different name. Gartner's research into agentic AI says the quiet part out loud: most agentic projects are 'early stage experiments or proof of concepts that are mostly driven by hype and are often misapplied,' and 'many use cases positioned as agentic today don't require agentic implementations.' A pilot that impresses the leadership offsite but moves no number is the corporate version of a chatbot with great retention and no revenue.
2. The 40% Cull Has a Guest List
Gartner expects more than 40% of agentic AI projects to be cancelled by the end of 2027 — and estimates that of the thousands of vendors calling themselves agentic, only about 130 are real. The rest are 'agent washing': a chatbot, an RPA script or an assistant relabelled with the word of the year. Here is how to tell which side of that line a tool is on in under a minute. Ask what it does when nobody is prompting it. An assistant waits. An agent has a standing instruction, a schedule, a set of tools and a definition of done — and reports back with what it did, not with a text box.
3. Buy the Outcome, Not the Demo
This is the design principle underneath every AI Employee in Kavanah. You do not hire one to chat; you hire one to own a workflow. 'Every weekday at 7am, triage new requests, assign owners by skill profile, flag anything due this week with no update, and send the digest.' That is a job. It runs on a schedule, inside a workspace where the tasks, the people, the time entries, the deals and the client threads already live, so the agent does not have to be told what the business looks like — it can see it. And when it finishes, there is a concrete artefact: tasks assigned, follow-ups sent, a report in the inbox. Not engagement. Output.
4. Make the Agent Prove Its ROI Every Month
The second Character.AI lesson is about cost discipline. Kavanah bills every model call against credits, at the price of the model that actually ran — and the chokepoint that makes the call is the thing that records the charge, so there is no quiet category of 'free' AI work that becomes a $45-a-day surprise. Each AI Employee's runs, tokens and actions are visible, and the workspace owner can cap what headless automation may spend in a day. If an agent is not paying for itself, you will see it in the ledger before you see it in the P&L. That is the opposite of a product that counts minutes of attention and hopes.
5. The Question for Your Q4 Budget
Before you renew or buy anything with 'AI' in the name this quarter, write one sentence: 'This tool exists to produce ___ by ___, and we will know it worked because ___.' If you cannot fill in the blanks, you are not buying a worker; you are buying a very expensive way to feel modern. Character.AI had the best technology in its category and could not fill in that sentence. The teams that win the next three years are the ones that refuse to deploy an agent until they can.
A billion dollars of valuation evaporated because the product had users but no job. Give your AI a job description before you give it a login.
