Infraestruturas de Portugal jumps with two feet into the AI agentic era

I saw on the cover of a portuguese financial newspaper a couple of weeks ago a piece about Infraestruturas de Portugal (the state-owned infrastructure public company) tendering a €1M contract for an AI tool called BIA - Bot de Inteligencia Artificial (Artificial Intelligence Bot) to be used to assist in their procurement processes for goods and services (but not public works). Intrigued, I chased the rabbit down the hole and got my hands on the tender documentation. 

The documentation unearthed a number of interesting things that are worth comenting on. In fact, so many that they do not fit into a single blogpost and warrant a full series of multiple blogposts. The first one will be about the choice of procedure and timeline, the second on contract object, the third on contract value, cost and award criteria. 

Usually, I do not cover substantive issues on the blog as I prefer to stay close to my sphere of competence but as it happens the subject matter in this case is close enough to my wheelhouse for me to comment on. In fact, one single post may not even be enough for that part. But first, the procedural side of things.

Choice of procedure

Infraestruturas de Portugal adopted an open procedure to tender this contract. This is the first questionable choice for a contract of this type. The company core business is that of managing and building (well, tendering construction contracts) for physical infrastructure. It is not technology and certainly not cutting edge technology. In short, it does not seem to have prior experience on AI systems, let alone AI bots or AI agents (the wording is used interchangeably throughout the documentation). So, why use an open procedure that pressuposes a good enough level of understanding of what needs to be done and the best solution to do so? 

It does not make much sense to use an open procedure in a situation like this when both competitive dialogue, innovation partnership and even the competitive procedure with negotiation (longer shot) are all available. The open procedure does not seem to be the right tool for this particular job and Infraestruturas de Portugal has painted itself into a corner here.

The counterargument here is that perhaps they did some market engagement before launching the tender, but I could find no evidence of that on the national tender portal. And if they simply rang a company and asked for advice, well, then that is not really market engagement and more of a red flag (more about red flags later).

All in all, this does not look to me as a great start and it is unsurprising there was a challenge by an economic operator about the procedure....which may or may not mean the whole thing has already stopped in its tracks.

Timeline (procedure)

The contracting authority adopted a very short timeline for both the procedure and contract performance. For the procedure here's the timeline summary:

• Initial notice: 04/02
• Tender submission deadline: 09/03
• Amendment to tender submission deadline notice: 10/03
• Amended submission deadline: 16/03

The first thing to point out here is that for a fairly complex contract, economic operators were originally given a grand total of 30 days to submit their tenders. This is the legal minimum...and not really recommended for complex contracts like this one. In fact, I would describe this as an an example of a disproportionate requirement and another red flag. The use of short submission deadlines for discriminatory purposes of tilting the playing field in favour of a particular supplier has been well established in the literature. The fact the submission deadline was amended is a sign that the original deadline was disproportionate for the market to respond.

But that is not the only problem here. See anything funny with those dates? The amendment notice was published only after the original submission deadline. Funny enough, said notice has a date of 06/03 on its text but that is irrelevant. What is relevant is that the actual publication date which ocurred beyond the original submission deadline. We will never know how many economic operators that did not submit a bid would have done so if they knew they had 25% more time. This is yet another red flag with the procedure.

Timeline (performance)

Then we have the timeline for contract performance. Infraestruturas de Portugal is looking to obtain the first version of the tool by the end of June and the final version by the end of July. For a project of this complexity the contractor will have to deliver the solution in a very short timeframe. Again, this looks like a red flag and it means that only supplier(s) with existing solutions are able to compete. But since we are talking about something new and fairly cutting edge, how many can do so? Or are they betting on a vibecoded solution done quickly?

There can be a more prosaic explanation here though, and that is money or, better said, the source of the money to pay for the contract. The contract notice discloses the use of EU funds to pay for this contract. My gut feeling is that these are probably coming from the Recovery and Resiliency Facility, whose deadline for contracts to qualify for is the end of August. I think the timelines are too close for this not be a possible explanation for the decisions on choice of procedure, duration and performance. Nonetheless, the red flags remain. And as we say in Portugal "rush is never a good counsel."

Tomorrow we will be looking at the object of the contract. What is Infraestruturas de Portugal actually looking to buy here?