When a bid looks too good to be true, AI might be the ghostwriter
As generative AI becomes more accessible and powerful, suppliers across sectors, from micro-SMEs to global contractors, are turning to tools like ChatGPT, Claude, and Gemini to prepare their bids. It’s easy to understand why: these tools promise fast, fluent, and often dazzlingly polished content. But beneath the slick presentation lies a growing challenge for public procurement evaluation panels: are we assessing the supplier’s capability, or their AI’s capability to write about it?
The Style Over Substance
An increasing number of bids are landing on evaluators’ desks that read like they were composed by a novelist trained in procurement strategy. Answers are fluid, structured, and full of confidence, but often light on specific, contextual substance. The risk is clear: AI can help a supplier sound far more competent than they are, using generalised language that creates a convincing illusion of expertise.
While there’s nothing wrong with help crafting a more articulate bid, the danger emerges when suppliers rely entirely on AI to generate answers about work they haven’t done, don’t understand, or can’t deliver. It creates a new kind of delivery risk, one not from incompetence, but from illusion.
Fake CVs, Phantom Experts, and Synthetic Teams
One particularly alarming development is the rise of AI-generated CVs and team bios. With tools capable of inventing fictional professionals who appear entirely real, complete with detailed project experience, qualifications, and references, evaluation teams face a new frontier in due diligence.
Public buyers have already encountered cases where key personnel listed in tenders either did not exist or were unaware they were being proposed. The scale and speed at which AI can produce these fakes increases the challenge exponentially.
What can be done?
- Require validation of key personnel (e.g. LinkedIn, signed consent forms).
- Cross-check project claims with known data (e.g. Contracting Authorities’ prior records).
- Implement structured interviews for named individuals as part of the evaluation.
The integrity of public procurement depends not only on price and compliance—but on the truthfulness of what’s submitted.
The Capability Gap: When AI Knows More Than the Bidder
Generative AI can produce detailed technical answers even if the bidder has no background in the topic. This creates a dangerous capability gap: a situation where the bid is judged favourably, but the supplier lacks the competence to deliver it.
Especially in construction and engineering sectors, this can be catastrophic. A compelling AI-generated method statement doesn’t equal capacity on the ground. Evaluation panels must not fall into the trap of rewarding style over track record.
For complex or higher-risk projects, additional evaluation stages, such as supplier presentations, clarification rounds, or live scenario testing, can help bridge the gap between words and reality.
More Supplier Presentations, Stronger Exit Clauses
As AI enables more suppliers to produce bids that look deliverable, procurement teams may need to rely more on in-person supplier presentations. These sessions allow evaluators to test understanding, probe inconsistencies, and watch how well suppliers respond without AI prompts in the background.
Similarly, contracts may need stronger exit mechanisms in place to protect against performance shortfalls resulting from AI-inflated bids. This could include milestone-based penalties, tighter KPIs, or suspension clauses triggered by misrepresentation.
But it must be done carefully: too many hurdles may scare off genuine, smaller suppliers who use AI simply to express themselves more clearly, not to deceive.
Fairness and Guidance – Not Fear
It’s tempting to consider banning AI-generated content altogether, but such a move would likely backfire. Assistive AI tools level the field for linguistically challenged suppliers, non-native speakers, and SMEs without professional bid teams. A fairer approach is to provide clear guidance:
- Declare whether assistive tools were used.
- Ensure key facts are independently verifiable.
- Accept AI support for clarity, but not for fabrication.
Don’t Evaluate the Syntax, Evaluate the Essence
Ultimately, a bid’s value lies not in how well it reads, but in how well it reflects real, capable delivery. AI will continue to sharpen the presentation of bids, but evaluation panels must sharpen their scrutiny in tandem.
Human intelligence, not artificial, is what makes public procurement resilient, fair, and trusted.
Sources:
OECD, Risks and Challenges of AI in Procurement (2023)
European Commission – AI Watch: Generative AI and Procurement (2024)
