How public procurement can harness artificial intelligence without losing human judgment.
In a world where bids arrive in volumes that would drown any sane evaluator, artificial intelligence (AI) is not just helpful, it’s becoming essential. But while the allure of automation is strong, it must be tempered with an unshakeable commitment to due diligence, risk management, and, frankly, good old-fashioned common sense.
The Efficiency Argument: Not Missing a Beat… or a Document
Tendering, especially in construction, generates hundreds of pages of submissions, annexes, technical drawings, CVs, insurance records, method statements, ESG credentials, and more. Human evaluation teams, often under tight timelines, are expected to comb through this haystack in search of compliance needles. AI can help.
Modern AI tools, whether integrated into procurement platforms or used as standalone engines, can automatically detect missing documents, mislabelled attachments, or logical inconsistencies within a bid. This doesn’t just improve efficiency; it protects against legal challenge. Evaluators can rely on a systematised, repeatable logic to show that no bid was unfairly overlooked or misunderstood.
For lower-value or routine projects, where high scrutiny isn’t always proportionate, AI-driven completeness checks could enable smaller teams to manage more competitions with fewer errors. However, this presumes that the humans in the loop understand the outputs, and do not blindly trust them.
Inclusion at the Core: Leveling the Field with Assistive AI
AI has a lesser-known but powerful side effect: it promotes equity. Many public procurement bodies, especially across Ireland and the UK, are under increasing pressure to ensure open and inclusive access to opportunities, particularly for SMEs, social enterprises, and first-time bidders.
For those with dyslexia, cognitive processing challenges, or who are bidding in a non-native language, AI-powered writing aids like Grammarly, ChatGPT or Google’s Gemini offer a critical support layer. These tools can help structure responses, improve tone, or simplify dense technical information.
The use of assistive AI should not be seen as a shortcut or a deceit, it is, in many cases, a compensatory technology. Much like a wheelchair ramp at a government building, it ensures the door to competition is open to all. Evaluation teams must be trained to distinguish between genuine use of AI as an enabler versus superficial, AI-generated content that masks capability gaps.
Risk-Based Approach: Not All AI Use is Created Equal
The presence of AI-generated text should not automatically disqualify a bidder, but it should trigger a thoughtful response. When bid evaluation software flags potential AI use (e.g., through watermarking or prompt residue), evaluators should apply a risk-based lens.
For example:
- Is the bid for a €5 million bridge, or a €20,000 maintenance project?
- Does the bidder show a grasp of local context, regulations, and technical specifics?
- Is there a track record of similar delivery?
- Are technical and financial proposals internally coherent?
The challenge lies in determining whether AI has been used to assist understanding or to replace it. While well-written responses are always welcomed, they should not distract from verifying actual delivery capacity.
A Note on Standards and Exit Clauses
AI is also subtly changing what contracts need to look like. If a contractor wins based on an AI-enhanced bid but fails to perform, authorities must be prepared to act quickly and proportionately.
Currently, most public contracts assume human authorship and performance. Clauses designed to address performance defaults may not be adequate when faced with failure due to AI-induced overpromising, reliance on synthetic team members, or ghost subcontractors.
Tightening contract exit mechanisms, such as performance bonds, milestone reviews, or dynamic penalties, can offer a safeguard. However, these must be balanced against their potential to inflate bid prices or deter legitimate SMEs who might rely on AI for efficiency.
Bottom Line: Use It, Don’t Worship It
Evaluation teams should embrace AI for what it is: an intelligent, assistive tool, not an oracle. It can help process large volumes of data, improve accuracy, and support a more inclusive and equitable procurement process. But it cannot replace informed judgement, domain expertise, or ethical oversight.
A future-fit evaluation framework should blend the best of AI’s processing power with the irreplaceable intuition of trained professionals.
Sources:
European Commission, Ethics Guidelines for Trustworthy AI (2020)
UK Government, Guidelines on Responsible AI in the Public Sector (2022)
World Economic Forum, AI Procurement in a Box Toolkit (2020)
