Cato has raised €6 million for artificial-intelligence software that automates much of the work companies perform when competing for public contracts, targeting a procurement market where tender notices have become easier to publish digitally without making the underlying job of interpreting requirements and preparing compliant bids much easier.

Keen Venture Partners led the seed round, with Italian Founders Fund, Vento, Heartfelt, Moonstone, BHeroes, Alecla7, Nova Venture, and several angel investors also participating. The financing follows an earlier pre-seed and takes the Milan company’s total funding to €7.6 million.

Cato operates on the supplier side of public procurement. Its software searches tender sources, ranks opportunities against a customer’s profile, analyses documents, extracts requirements and evaluation criteria, checks apparent eligibility, and assists with administrative and technical material using information provided by the bidder.

The workflow is well suited to automation because much of the work involves repetitive reading, comparison, and document preparation. It is also unusually unforgiving because a fabricated certification, incorrect financial statement, or inaccurate description of past performance can invalidate a bid or create wider legal and commercial problems.

Digital procurement creates material for agents

European governments purchase goods and services on a scale that makes public procurement one of the continent’s largest business markets, ranging from construction and healthcare equipment to software, maintenance, and professional services. The increasing digitisation of tender notices and supporting documents has created a large pool of structured and semi-structured information that software can search more effectively than earlier paper-based processes allowed.

Digitisation has not removed the administrative burden on bidders, however. Companies still need to decide which opportunities are worth pursuing, establish whether they satisfy formal requirements, interpret long specifications, coordinate internal contributors, manage deadlines, and assemble evidence without making a mistake that disqualifies the submission.

Cato says its system monitors more than 27,000 sources and uses information about each customer to filter irrelevant opportunities before helping with document analysis and preparation. That filtering may prove as valuable as generation because a company saves little time if cheaper bid production encourages it to chase contracts that it is unlikely to win.

The final submission remains with the customer, an important boundary in a process carrying legal consequences. Agentic software can take on more of the workflow, but accountability remains with the company signing the bid and ultimately delivering the contract.

Traceability becomes a product requirement

Procurement provides a useful test for AI agents because the documents involved are dense and repetitive while the output still needs evidence. Models can extract, compare, summarise, and draft quickly, yet the user must be able to establish where each factual claim came from and whether it accurately represents the organisation.

Cato says material generated through its platform is linked to source information, while customers retain control over submission. The company also states that customer and tender data is not used to train general-purpose models, and it advertises EU-based data storage alongside ISO 27001-certified development processes.

Those controls become more important as software moves from drafting assistance towards autonomous completion of tasks. Procurement teams handle pricing, customer references, certificates, staff information, strategic documents, and other material that companies may be unwilling or unable to expose to loosely governed model providers.

Security and evidence therefore sit inside the product proposition rather than alongside it. An agent capable of producing a complete tender response is useful only if a customer can see which information it relied on, correct errors, control sensitive data, and understand what the system actually did.

Cheaper bids can create new problems

Reducing the cost of preparing tenders could help smaller companies compete for more public work, particularly where administrative overhead makes opportunities uneconomic to pursue. A business with a small sales team may be able to assess more contracts when software handles the repetitive first pass through documents.

More bids do not automatically improve procurement, however. If automated drafting encourages suppliers to respond indiscriminately, contracting authorities may receive larger volumes of generic submissions and spend more time assessing offers that are poorly matched to the requirement.

Useful automation therefore needs to improve the decision not to bid as much as the decision to proceed. Eligibility checking, opportunity ranking, and assessment of commercial fit can prevent AI from simply lowering the cost of producing paperwork that should never have been submitted.

The underlying public-procurement system also retains structural problems that supplier-side AI cannot solve. Fragmented portals, complicated specifications, long qualification processes, cautious buying behaviour, and inconsistent payment practices can deter companies regardless of how efficiently software reads the documentation.

European expansion will test portability

Cato says it has processed thousands of tenders and serves customers across sectors including medical devices, construction, IT, and services. Those adoption figures come from the company and its investor, but they indicate that the product has moved beyond an early technical demonstration.

The new capital will support product development, hiring, and expansion beyond Italy, including work with multinational customers. That will test how portable the current system is across European procurement regimes because document processing can be reused while legal terminology, portals, procedures, and qualification rules vary nationally.

Public tenders are consequently an attractive environment for agentic AI without being an easy one. The documents are abundant, the repetitive work is real, and the economic value can be large, but errors remain attributable to the bidder rather than the model.

Cato’s development will show whether procurement can become a durable category for enterprise agents rather than merely another source of documents for generative AI to summarise. The more autonomy the software gains, the more its commercial value will depend on preserving evidence, security, and accountability alongside speed.