commercetools and Mirion Launch the B2B Intake Agent to Turn Emails and PDFs Into Quotes in Minutes
commercetools co-designed the B2B Intake Agent with manufacturer Mirion Technologies to convert unstructured emails, PDFs, and spreadsheets into editable quotes and carts in minutes. Here is how it works and what it means.
commercetools and Mirion Launch the B2B Intake Agent to Turn Emails and PDFs Into Quotes in Minutes
Key Takeaways
Key Takeaways
- commercetools co-designed a new AI agent called the B2B Intake Agent with nuclear-measurement manufacturer Mirion Technologies to process unstructured order requests
- It matches SKUs and quantities scattered across emails, PDFs, and spreadsheets to the catalog and generates editable quotes and carts tied to the right account and pricing in minutes
- The output stays a draft that sales reps review and edit, keeping the human as the decision point, a restrained design that fits accuracy-critical B2B transactions
What Was Announced
commercetools and Mirion launch the B2B Intake Agent, turning emails, PDFs, and spreadsheets into ready-to-action quotes and carts in minutes.
www.demandgenreport.comOn June 16, 2026, composable commerce provider commercetools announced a new AI agent called the B2B Intake Agent. The capability automatically converts order requests that arrive in unstructured formats, such as emails, PDFs, and spreadsheets, into quotes and carts.
What stands out is that this feature was not designed in a vacuum. It was built together with Mirion Technologies, a manufacturer of nuclear measurement and detection systems. The requirements of a company that handles complex orders every day were folded into development from the earliest stage. commercetools frames this arrangement as a co-design partnership built alongside a real enterprise customer.
How the B2B Intake Agent Processes Unstructured Orders
The heart of this story is what the agent actually converts, and how. B2B order requests do not always arrive through a tidy ecommerce cart screen. They come as the body of an email written by a buyer, an order form attached as a PDF, a line-item table in Excel, Word, or CSV. The formats vary as widely as the number of accounts.
The agent begins by ingesting these multi-format files. It extracts the substance of the order, such as SKUs and quantities, and automatically matches them to the commercetools product catalog. Using the sender's email address as a clue, it resolves the account (business unit) already in the system and ties the request to that customer's pricing and terms. This is the work of aligning scattered information to the correct account and pricing structure.
The output is a draft quote and cart that a sales rep can use directly. The important point is that the agent does not finalize the transaction on its own. According to commercetools, the generated quote or cart can be freely edited by the rep, or frozen when needed. Final accuracy and customer-specific adjustments remain with the human side.
The technical positioning is equally clear. This is not a screen the buyer touches but a seller-side agent embedded in internal sales operations. It is API-first, synchronizing in real time with upstream and downstream systems such as ERPs, CRMs, and procurement workflows. Integration with the CS platform Zendesk via API is one cited example. The design aims to slot into existing workflows without a separate orchestration layer.
The Weight of Manual Work Behind the 70% Figure
Why is there demand for this capability? commercetools explains the backdrop through how sales time is spent. According to a DocuSign study cited in the release, sales reps spend roughly 70% of their time on administrative tasks. Reading incoming orders and rekeying SKUs, prices, and quantities into backend systems slows response times and drives up costs.
On its own blog, commercetools also offers an estimate that puts this burden into money. For an organization processing 1,000 orders of 50 lines each per month, data entry alone consumes about 750 hours a year, which works out to roughly 600,000 euros in annual labor costs. The figure is a model case, but it shows how manual entry of unstructured orders can pile up to a scale that is hard to ignore.
The headline benefit is a shorter quote lead time. A quote reply that used to take hours or days can be compressed into minutes. Faster responses, in commercetools' telling, feed conversion, average order value, and retention.
What the Co-Design With Mirion Signals
The choice of Mirion as a partner carries meaning. The company deals in nuclear measurement and detection, a product line that is complex and highly specialized. Orders involve specific requirements and conditions, and mistakes are hard to tolerate.
Matthew Maddox, VP of Digital Commerce at Mirion, put it this way in the announcement. "For manufacturers like Mirion, speed matters, but accuracy matters just as much." He added that he sees strong potential for AI to simplify order intake, improve responsiveness, and let teams focus on solving customer needs. Placing accuracy concerns openly alongside speed reflects a manufacturer's operational instinct.
The announcement sits on the extension of commercetools' broader AI strategy. In January 2026, the company released AgenticLift, a standalone product that lets enterprises join AI shopping without replatforming. In June, it introduced the concept of autonomous commerce, where AI executes operational decisions in real time. The B2B Intake Agent reads as an implementation that grounds that autonomy vision in the concrete task of order intake. Shiri Mosenzon-Erez, CPO at commercetools, framed the problem by noting that at many B2B businesses, talented sales and service teams spend too much time translating order requests instead of serving customers.
The Open Questions in Adoption
That said, handing B2B order intake to AI is not all smooth sailing. The reservations deserve equal attention.
One is how to guarantee accuracy. The very fact that this agent adopts a design that sends drafts to a human for review can be read as caution about full automation. In orders for complex products, a single mismatched SKU can erode trust in the transaction. Every organization needs to run a process that verifies the output rather than taking it at face value. The more the human review step remains, the smaller the labor savings become relative to the theoretical maximum.
Another is the plumbing. Industry discussion notes that legacy systems constrain integration and flexibility at more than 60% of organizations. AI struggles to deliver value on top of fragmented systems and disconnected workflows. Even with an API-first design, if internal data and processes are siloed, the catalog and pricing alignment that the agent depends on can break down.
Availability timing, pricing, and general-availability terms are currently undisclosed. This announcement signals the result and direction of a co-design effort, and the details of which size of company can use it at what cost await further updates.
Conclusion
What the B2B Intake Agent illuminates is how the agentic commerce conversation has widened from AI acting for the buyer to AI handling the seller's order intake. Rather than flashy autonomous buying, the pragmatic insight is that the dull work of rekeying orders from emails and PDFs is exactly where AI applies most cleanly.
At the same time, this case shows a restrained design philosophy that keeps human review for complex B2B transactions. In a domain where neither speed nor accuracy can be sacrificed, positioning the agent as a draft generator rather than the decision point is one plausible answer for now. The next question is whether this co-design pattern spreads beyond a single customer, terms of availability included, into a form many enterprises can use.
