Adobe Is Buying Its AI Stack, Not Building

Count the days between Adobe promising an AI coworker and Adobe buying one. It comes to 135. On April 20 at Summit, Adobe unveiled CX Enterprise Coworker, a system that takes a business objective and turns it into multi-step action, and said it would be generally available in the coming months. On September 2, it bought a twelve-month-old startup whose agents already did exactly that, and switched the product off. Co-founder Dhruv Jaglan announced the acquisition of Rilo himself, telling users his team would help them move their workflows out. Adobe has published nothing about it, and financial terms have not been disclosed.

Start with what that team actually built, because it explains what a company Adobe’s size could not assemble in five months. Rilo’s agents let non-technical users describe a workflow in plain English and then execute it across hundreds of tools, covering competitor intelligence, content repurposing and distribution, and sales call analysis. Adobe already sells content generation at enterprise scale. What it did not have was the connective tissue between a business request and the sequence of actions that satisfies it.

The shutdown has been treated as a footnote, and it is the most informative fact in the deal. Everyone who built a workflow inside Rilo now rebuilds it somewhere else. A platform genuinely running a marketing function for paying customers does not get switched off at close. It gets migrated, or kept alive alongside, because customers object loudly. Shutting it down is what happens when the customer relationships were never the asset.

Read as a sequence, the past five months look less like opportunism than like an admission. Eight days after Summit, Adobe completed the Semrush acquisition for $1.87 billion, primarily in cash. Of that purchase price, $1.25 billion was recorded as goodwill, which the quarterly filing attributes primarily to the assembled workforce and expected synergies. Adobe’s own accounting says the people were the asset. Enterprise software companies are slow at this by nature, and buying people who have done it once is faster than briefing a team that has not.

This is not an isolated week. Optimizely announced Virtual Teammates at Opticon on August 31, role-specific AI coworkers built to hold standing responsibilities and retain context instead of being prompted task by task. Its own study of more than 2,000 B2B marketing leaders found 81% personally switch between two or more disconnected AI tools every week. The category is converging on one claim, which is that the software does the work rather than assisting with it.

What does this mean for a team with no Adobe contract and no plans for one? The roadmap for the tools you do use is increasingly set in acquisition meetings rather than product ones, and the likeliest to be absorbed are the young, single-purpose ones your ops team adopted because they were fast to stand up.

There is a second signal worth reading here. Buyers already sense the instability: G2’s 2026 Buyer Behavior Report found that seven in ten buyers say the pace of AI innovation is pushing them toward shorter contracts. That instinct now has a worked example.

One caution before anyone treats this as category validation. More than 10,000 people had built with Rilo within months of launch, according to Jaglan, which is evidence of interest rather than evidence of a durable business. There is no disclosed revenue, paying-customer count or retention data. Adobe’s acquisition confirms that the technology and team were strategically valuable to Adobe. It does not establish that the market was paying for the product at scale.

The practical move is an audit rather than a reaction. List the workflows that live inside a vendor younger than two years, and for each one answer two questions: can you export the logic, or only the output, and if the product disappeared next quarter, how many days of rebuilding would that cost. Then put sunset and change-of-control language in the contract while you still have leverage.

Adobe bought twelve months of someone else’s engineering and turned the product off on the way in. For anyone building on a young tool, the lesson is that your workflow can be an asset to an acquirer and a liability to you at the same time.

What does this mean for a team with no Adobe contract and no plans for one? The roadmap for the tools you do use is increasingly set in acquisition meetings rather than product ones, and the likeliest to be absorbed are the young, single-purpose ones your ops team adopted because they were fast to stand up.

There is a second signal worth reading here. Buyers already sense the instability: G2’s 2026 Buyer Behavior Report found that seven in ten buyers say the pace of AI innovation is pushing them toward shorter contracts. That instinct now has a worked example.

One caution before anyone treats this as category validation. More than 10,000 people had built with Rilo within months of launch, according to Jaglan, which is evidence of interest rather than evidence of a durable business. There is no disclosed revenue, paying-customer count or retention data. Adobe’s acquisition confirms that the technology and team were strategically valuable to Adobe. It does not establish that the market was paying for the product at scale.

The practical move is an audit rather than a reaction. List the workflows that live inside a vendor younger than two years, and for each one answer two questions: can you export the logic, or only the output, and if the product disappeared next quarter, how many days of rebuilding would that cost. Then put sunset and change-of-control language in the contract while you still have leverage.

Adobe bought twelve months of someone else’s engineering and turned the product off on the way in. For anyone building on a young tool, the lesson is that your workflow can be an asset to an acquirer and a liability to you at the same time.