AI can help bid teams analyse tender documents, organise existing content and complete early drafts more efficiently. However, increased use brings greater scrutiny. Public sector buyers want to understand whether AI has influenced a tender, what information has entered the system and who has verified the final submission.
That doesn’t mean suppliers should stop using AI. It means they need to use it deliberately, protect buyer information and remain accountable for every claim they submit. At Thornton & Lowe, we see this as an important part of professional bid governance, rather than another box to tick before the deadline.
PPN 017 Makes AI Use More Visible
The Cabinet Office published PPN 017: Improving Transparency of AI Use in Procurement in February 2025. It applies directly to central government departments, their executive agencies and non-departmental public bodies. Other public sector contracting authorities may also choose to follow its approach.
PPN 017 reflects terminology introduced by the Procurement Act 2023 and Procurement Regulations 2024. Its provisions apply to procurements commenced on or after 24 February 2025. Procurements started before that date remain subject to the previous PPN 02/24.
For suppliers monitoring wider procurement changes, the main point is simple: buyers are being encouraged to understand how AI is being used in both tender preparation and contract delivery.
The guidance includes optional disclosure questions that buyers can add to their procurement documents. These questions can ask suppliers:
- whether AI or machine learning was used to create any part of the tender response
- whether AI-generated content has been checked and verified for accuracy
- whether AI or machine learning will form part of the proposed product or service
These are example questions, rather than a universal declaration that appears in every tender. Suppliers must therefore read the specific procurement documents carefully instead of assuming that the same disclosure rules will apply to every opportunity.
Some Procurements Carry Extra National Security Considerations
PPN 017 also flags that certain procurements involve additional national security risk connected to AI use. Where a contracting authority identifies this kind of concern, its commercial team is expected to engage Information Assurance and Security colleagues before the procurement even launches, so that proportionate safeguards are built in from the outset.
For suppliers, this means the level of scrutiny around AI use isn't uniform across the public sector. A framework opportunity in a sensitive sector, such as defence, critical infrastructure or central government, may carry a materially different risk profile to a routine services contract. Suppliers bidding into these areas should expect closer questioning about how AI tools are used, what data they touch and where that data is processed, and should have answers ready rather than reactive.
This sits alongside the wider national security procurement guidance under PPN 025, which sets out how buyers assess and mitigate national security risk more broadly.
AI Use Is Allowed, but Suppliers Remain Accountable
PPN 017 explicitly recognises that AI may help suppliers bid for more public contracts. It also confirms that using AI during the commercial process is not prohibited.
As such, a supplier shouldn’t assume that declaring responsible AI use will automatically damage its score. Buyers are more likely to be concerned by vague answers, weak controls or a supplier that cannot explain how its tender was produced.
Used carefully, AI in bid writing can support activities such as:
- analysing the requirements and evaluation criteria
- producing an initial response structure
- identifying possible gaps in a draft
- summarising approved internal material
- improving clarity and consistency
- checking whether a response addresses every part of the question
However, the supplier remains responsible for the submission. That includes every statistic, case study, policy reference, staffing commitment, delivery timescale and performance measure.
An AI platform cannot approve a contractual commitment on behalf of your business. It also cannot confirm that an attractive-sounding claim reflects your genuine capacity to deliver.
What Should an AI Disclosure Include?
Where a buyer asks whether AI has been used, an effective response should be accurate and proportionate.
A limited use of AI for proofreading may only need a brief explanation. A tender substantially drafted or analysed using generative AI may require more detail. Suppliers should follow the wording of the question and provide enough information for the buyer to understand the level of risk.
A credible declaration could cover five areas.
1. Which Tools Were Used
Identify the platform or type of technology where the buyer requests this information. Avoid referring vaguely to “digital tools” if generative AI has made a material contribution.
2. How AI Supported the Response
Explain the tasks involved. For example, AI may have helped to structure an answer, compare a draft against the specification or improve the clarity of existing human-written content.
The explanation should distinguish these activities from generating new evidence, case studies or delivery commitments.
3. What Information Entered the System
Confirm how you controlled the information provided to the tool. This is particularly important where the tender documents contain confidential, commercially sensitive or personal information.
4. How the Output Was Verified
Describe the human review undertaken. Strong tender writing tips remain relevant here: answer the exact question, follow the instructions, use evidence and check every claim before submission.
5. Who Approved the Final Tender
Accountability should sit with a named role or suitably authorised team. The final reviewer must be able to confirm that the response is truthful, compliant and deliverable.
Keeping a simple internal record of where AI was used can make these declarations easier. It is far harder to reconstruct the process shortly before submission, particularly when several contributors have used different tools.
Service Delivery Needs Separate Consideration
AI disclosure isn’t limited to the creation of the bid. One of the example questions in PPN 017 asks whether AI or machine learning will be used as part of the products or services supplied under the contract. The guidance notes that AI can be incorporated into services that aren’t primarily marketed as AI products, such as transcription or translation functionality within video-conferencing platforms.
Suppliers should review their full delivery model, including third-party software and subcontractors.
A buyer may reasonably want to know:
- what function the AI performs
- whether it supports or makes decisions
- what information it processes
- whether personal or sensitive data is involved
- where the information is stored
- whether customer data is used to train a model
- which third parties have access
- how outputs are monitored
- how errors can be challenged or corrected
- what happens following a security incident
This can affect suppliers far beyond the technology sector. AI may be present in customer service systems, recruitment software, scheduling platforms, monitoring tools, document processing, fraud detection or operational reporting.
Procurement teams also need proportionate questions. A basic spelling assistant presents a different risk from an AI system processing service-user data or influencing decisions about access to a public service.
Protect Buyer Data Before Using Any AI Tool
One of the clearest warnings in PPN 017 concerns confidential contracting authority information. Buyers are advised to consider controls preventing non-public information from being used as training data for AI systems.
Uploading an entire tender pack to a public AI platform without understanding its terms, data handling or retention settings can create an avoidable risk.
Before using a tool, suppliers should establish:
- whether it is approved for business use
- how prompts and uploaded documents are stored
- whether information may be used for model training
- how long the provider retains data
- where the data is processed
- whether access is restricted to authorised users
- whether an appropriate contract or data processing agreement exists
- how information can be removed
The Information Commissioner’s Office provides detailed guidance on AI and data protection, including how data protection principles apply when AI systems process personal information.
This should sit alongside your wider approach to cyber security and resilience. Buyers may examine the AI platform itself, but they can also look at access controls, staff training, incident management, subcontractor assurance and business continuity.
Human Oversight Must Be More Than Proofreading
PPN 017’s example disclosure questions ask suppliers to confirm that AI-generated content has been checked and verified for accuracy.
A quick read-through isn’t enough. Meaningful verification requires the reviewer to test the response against authoritative internal information. This includes checking:
- dates, figures and performance data
- certificates and accreditation details
- client names and case study outcomes
- staffing structures and availability
- proposed delivery methods
- implementation and mobilisation timescales
- KPIs and service levels
- legislation and policy references
- pricing assumptions
- contractual commitments
AI can generate content that is grammatically polished and structurally convincing while still being factually wrong. It may combine information from separate projects, misstate an accreditation or introduce a process that the delivery team has never used.
This is why human bid writing with AI tools remains the stronger model. Technology can increase speed and support quality control, but people provide the judgement, experience and organisational knowledge needed to produce a credible tender.
Create Controls Before the Next ITT Arrives
Suppliers shouldn’t wait for an AI disclosure question to establish their internal position.
A practical AI policy for bidding can be concise, but it should set clear boundaries. Your team needs to know which systems are approved, what information is prohibited and who is responsible for checking outputs.
A workable process could include:
- An approved-tools register
List the AI platforms that staff may use and any relevant licence, security or privacy conditions. - Information classification rules
Define what can and cannot be entered into generative AI tools. - Permitted bidding activities
Explain where AI may assist, such as analysis, structure or review. - Prohibited activities
Prevent the invention of case studies, qualifications, statistics or delivery commitments. - Human review responsibilities
Allocate checks to people with the knowledge and authority to approve the content. - Disclosure records
Record material uses of AI so the tender declaration can be completed accurately. - Subcontractor requirements
Ask delivery partners how they use AI and what systems may affect contract delivery. - Training and monitoring
Review the policy regularly as tools, buyer expectations and regulatory guidance develop.
The National Cyber Security Centre’s Guidelines for Secure AI System Development recommend treating security as a requirement throughout the AI lifecycle. The guidance covers secure design, development, deployment, operation and maintenance, including supply chain security, incident management, logging and monitoring.
Your Tender Library Is an AI Control
A strong content library reduces the risk of poor AI-generated answers. When bid teams start from verified company information, current case studies and approved policies, AI can help retrieve and organise genuine evidence. When they begin with an empty prompt, the system has more room to produce generic or inaccurate content.
Effective tender library maintenance should cover version control, ownership and review dates. Content relating to certificates, insurance, policies, staff, case studies and performance data must be updated before it is reused.
Purpose-built AI-powered bidding software can also provide more control than an unstructured collection of documents and standalone public tools. Thornton & Lowe’s Tender Library helps teams store, search and adapt established content, combining technology with a human-led bidding process.
The software won’t make an unsuitable claim true or turn a weak service into a strong one. It can, however, make approved information easier to find and give the bid team more time for tailoring, evidence and review.
Common Questions About AI in Public Procurement
Must suppliers always disclose AI use?
No blanket declaration applies to every public procurement. PPN 017 provides optional questions that buyers can include. Suppliers must follow the requirements and wording of each tender.
Where a declaration is requested, the answer should be complete and accurate. Deliberately hiding material AI use could undermine buyer confidence and create concerns about the reliability of the wider submission.
Will using AI reduce our tender score?
The example questions in PPN 017 are intended to be used for information and not scored. However, buyers may introduce additional scored questions where AI use is relevant to the contract, provided the criteria and assessment approach are clearly set out.
The quality of your governance may therefore matter more than the fact that AI was used.
Does proofreading software count as AI?
Potentially. Many everyday business platforms now include AI-supported editing, summarising or search features. Read the buyer’s definition and wording carefully.
Where there is doubt, a brief and proportionate explanation may be safer than an answer that appears incomplete.
What if a subcontractor uses AI?
The lead supplier should understand how AI is used throughout its proposed supply chain. This is especially important where a subcontractor processes data, supplies software, communicates with service users or influences operational decisions.
Can we submit AI-generated content without reviewing it?
You shouldn’t. Even where the tender doesn’t contain an AI disclosure question, the supplier remains accountable for its submission. Every response should be checked by someone who understands both the requirement and the proposed solution.
Use AI Without Losing Buyer Confidence
AI can improve the speed and organisation of your bidding process. It can’t take responsibility for your promises, protect data without suitable controls or decide whether a response reflects what your organisation can genuinely deliver.
The strongest approach combines clear AI governance, controlled source material and experienced human review. That gives buyers the transparency they need while helping your team work more efficiently.
Thornton & Lowe combines expert bid writing services with practical bid management support and purpose-built technology. Whether you need help with a live tender, stronger internal controls or a better way to manage approved content, we can help you build a bidding process that is faster, more consistent and ready for greater scrutiny.
