How to reduce risk
when preparing a bid

The documents for an international tender required the bidder to supply a brand-new pickup truck with insurance and fuel for twelve months, two workstations and survey equipment. The requirement sat in the additional information section. Outside the terms of reference, outside the price schedule: in a section almost nobody cross-checks against the budget.

Costing that line item came to 16% of the bid amount.

A proposal that reaches the evaluation table without it competes with a number missing one sixth of its value. And the tender package was in order: the requirement was written out in full, in the place nobody looks.

It surfaced because the documents were read from end to end.

Most organizations already using artificial intelligence started with a general assistant: draft a document, summarize a meeting, answer a one-off question. Useful tools, but they rarely change how the firm works. The step change comes from turning a tool into an operating capability: defined processes, written decision criteria, clear ownership and validation mechanisms.

At DG Ingeniería we apply this to bid preparation. We do not use a single assistant that tries to solve everything. We built a team of AI-assisted specialists, and the architecture rests on one decision.

Each specialist covers a narrow slice of the work, and only that slice.

Some analyze tender documents and addenda. Others check qualification requirements. Others review consistency across forms, flag technical risks, or support the calculation models. And none of them decides: whether to bid, at what price, and what gets signed is always a decision taken by a person.

That separation is what makes it possible to sweep the entire package instead of sampling it, with consistent criteria, without engineering judgment thinning out as the surface grows. The tool speeds up the reading. What decides where to look, and what a finding means, is judgment.

Most companies have already experimented with artificial intelligence; very few turned it into an advantage. The Build for the Future study by BCG, covering 1,250 executives, found that only 5% capture substantial value. The figure usually left off the chart is the middle band: a 35% that is already scaling and starting to capture it.

A working paper by Humlum and Vestergaard, published by the NBER and covering 25,000 workers, explains why the rest stay out. Within the same occupations, controlled trials document gains above 15%, while spontaneous adoption returns 2.8%. The model is the same on both sides. What changes is whether someone designed how it gets used.

That design is engineering work, and it is the part most often skipped.