Akeneo and Unifai: Why a PIM Vendor Bought Its Supplier Intake AI

Akeneo bought the French AI start-up Unifai in September 2023 so that supplier data intake would be part of its own Product Cloud instead of a partner app. Three years later the result sits in Akeneo Supplier Data Manager (SDM), where AI classifies supplier rows, fills in missing attributes and suggests column mappings before anything reaches the PIM.

The deal is an early example of the pattern from Part 1 of our Trends in PIM series. That piece argues that supplier onboarding is turning into its own layer in front of the PIM, with its own vendors, portals and prices. Akeneo saw that layer forming and chose to own it. This article looks at what the September 2023 announcement turned into.

What Akeneo bought

Akeneo announced the acquisition on 19 September 2023. The price was not disclosed. Akeneo's French press release describes Unifai as a start-up founded in 2017 and incubated at Station F in Paris. Its co-founder and CEO, Jesse Créange, told Station F that the founders went full time in 2020.

According to Akeneo, Unifai handled three jobs in the collection and cleansing step:

Supplier data onboarding. Cleansing, normalising and enriching supplier files on their way into a PIM.
Supplier pricing integration. Turning supplier price files into feeds for both the ERP and the PIM.
Marketplace seller onboarding. Cleansing seller data on its way into a marketplace back office.

The two companies already knew each other. Créange told Station F that Unifai met Akeneo in early 2021, worked with it as a partner and signed a reseller agreement in February 2023. The French release adds that Unifai had the highest adoption rate of the AI apps for product experience in Akeneo's App Store. Akeneo was buying a tool its customers already used, together with the team behind it.

Why Akeneo said it bought

Akeneo framed AI for product data as three steps: collection and cleansing, then enrichment, then translation. It placed Unifai in the first step and called that step the most critical. Fred de Gombert, Akeneo's CEO and founder, put the argument this way in the release: "Magically generating product descriptions in an ecommerce site or marketplace without centralized, complete and accurate product attributes will only lead to failed product experiences, poor customer touch points, and ultimately a damaged brand image."

In other words, generative copy is only as good as the attributes underneath it, and those attributes mostly arrive in supplier files. Owning the intake step gave Akeneo a story for both halves.

The release also carried customer figures. Rexel, a joint customer, reported 80% less manual effort and three times faster time to market. Intersport reported a 50% cut in product data processing time. Akeneo said a typical customer could aim for 50% time savings with 95% accuracy in data onboarding. The French version of the same release says 98%. These are vendor figures, not independent measurements.

What shipped since

The Unifai technology went into SDM and, according to Akeneo, into the PIM itself:

October 2023. Akeneo launched SDM as the next version of its older Akeneo Onboarder. It was sold alongside Akeneo PIM, with Unifai's AI as an option on top.
March 2024. The Spring 2024 release said Akeneo had built Unifai use cases natively into Akeneo PIM and SDM. For SDM that meant automated mapping, normalisation, categorisation and enrichment of supplier data.
2025. Akeneo's help centre describes the SDM AI agent as running on pre-trained large language models that need no training data or feedback loops. It covers classification, extraction and mapping, and AI-assisted bulk edits in normalisation are in early access. Since July 2025, admins can choose which source attributes the AI reads and add custom prompts.
2026. In March, SDM started suggesting column mappings, showing only matches with a confidence score of at least 70%, for files with up to 2,000 columns. Suppliers can now also start a job by emailing a spreadsheet to the portal, which is still in beta. Since July, SDM rejects asset files at upload when they break the image or video rules set in Akeneo PIM.

The Unifai brand is gone. When we checked on 21 October 2026, en.unifai.fr redirected to akeneo.com.

The help centre is also specific about where people stay involved. Classification can run in an automated mode, where only rows the AI could not classify go to a person, or in a mode where every row needs checking. Extracted values with low confidence are flagged for review. One detail matters for anyone who treats the supplier's own data as evidence: when extraction is set to re-evaluate every attribute, SDM does not notify you when its output differs from the value the supplier sent.

Bought, built or separate

Akeneo's route is one of three that PIM buyers now see side by side.

Bought and folded in. Akeneo bought Unifai and turned its technology into features of its own intake product. It has since bought again: PricingHUB, which Akeneo runs as a dedicated business unit. Our piece on Akeneo and PricingHUB covers that second step beyond product content.
Built as a door into the PIM. inriver took the other route. Its Spring 2026 release adds a Content Onboarding API with mapping and validation at intake, plus an extraction path for documents with a review step before anything is committed.
Separate specialists. Onedot sells AI onboarding into ERP, PIM or shop systems as its own subscription. openProd calls itself the onboarding layer between supplier files and the PIM. SDM Portal, which has no connection to Akeneo's SDM, maps supplier files to Ergonode, Akeneo or Pimcore once and reuses the mapping for every new file.

Each route moves the same work to a different owner. With a PIM vendor's own intake tool, one vendor owns both the target model and the intake rules. SDM, for example, syncs its taxonomy from Akeneo PIM. The tool's limits are then that vendor's limits: Akeneo's help centre lists 30,000 products per AI classification or extraction job and recommends no more than 50 source attributes. A specialist tool can serve several PIMs and ERPs, but, as Part 1 notes, it adds a contract and another place where attribute definitions can drift away from the PIM model. The lines are blurring too: Akeneo now describes SDM as a way to import product data into any system used for product records, not only its own PIM.

Where the routes differ in practice

When teams compare built-in onboarding with a specialist intake tool, the differences tend to show up in four places.

Low-confidence values. Some tools hold uncertain values for a person and some write them through with a flag. Check which one each tool does, and whether you can set that per step.
Ownership of the target model. AI can suggest a mapping, but someone still has to own the attributes it maps to. Our Onedot session recap makes the same point about AI mapping in general.
What the vendor's numbers measure. Akeneo's own 2023 release gave two different accuracy figures for the same claim. A percentage without a dataset and a definition says little about your suppliers.
Your own supplier file. The fastest way to see the difference between routes is to run one messy file through each. Our guide to bringing your worst supplier file to the PIM demo describes how to set that up.

The Unifai deal did not settle whether intake belongs inside the PIM vendor or next to it. It showed that a major PIM vendor considered the step valuable enough to buy, and three years of SDM release notes show how that bought technology became a product of its own.

Sources

Diagnostic

Do you actually need a PIM?

Run the complexity index before you budget software or hire an SI.

Start assessment

Budget

Model a first-pass TCO

Translate catalog shape into a three-year cost range in under ten minutes.

Open cost calculator