Artificial intelligence is increasingly being used to support commercial activity. Businesses may use it to prepare first drafts of contracts, review agreements, summarise negotiations, identify clauses or manage large volumes of contractual information.
Used carefully, these tools may improve efficiency. However, producing contractual wording quickly is not the same as producing an agreement that properly protects the business.
Commercial contracts reflect the objectives of the parties, the risks each side has agreed to accept and the compromises reached during negotiation. They must also operate within an existing framework of legislation, regulation and industry-specific requirements.
AI can support parts of this process, but it cannot replace the legal and commercial judgement needed to decide whether a contract is suitable for a particular transaction.
A contract must reflect the commercial arrangement
AI tools can produce wording that resembles conventional contractual drafting. This can make an AI-generated agreement appear polished and legally credible.
The difficulty is that a contract may be well written without being commercially appropriate.
A limitation of liability clause, for example, cannot be assessed by wording alone. The appropriate position will depend on the value of the agreement, the consequences of a breach, the parties’ bargaining strength, available insurance and the risks each side is willing to accept.
The same applies to warranties, indemnities, termination rights, service levels and intellectual property provisions. Each clause must be considered in the context of the wider transaction.
AI is unlikely to understand that a particular obligation was central to the negotiations, that one party made a significant concession elsewhere or that a standard clause would create an unacceptable exposure.
Human involvement is therefore needed to consider what the business is trying to achieve, what could go wrong, who should carry the risk and what remedy would be required if the arrangement breaks down.
Without that analysis, a contract may describe the transaction while failing to provide the protection the business expects.
AI-generated terms may not be legally suitable
Businesses should not assume that contractual wording produced by AI is legally compliant or up to date.
The law applying to commercial arrangements may change through legislation, court decisions and regulatory guidance. Certain sectors also operate under detailed rules affecting how services are provided, how information is handled and which contractual terms can be enforced.
An AI tool may not have access to the latest legal developments. Even where it identifies the correct legislation, it may not understand how it applies to the circumstances of the agreement.
Consumer-facing contracts require particular care. Terms should be fair, transparent and presented in a way that allows the consumer to understand their effect. Wording that may be acceptable in a negotiated business-to-business agreement could be inappropriate in a consumer contract.
Regulated businesses may also need agreements to address specific obligations concerning oversight, record keeping, security or customer protection.
AI may help identify issues, but it should not be treated as confirmation that an agreement complies with every applicable requirement.
A convincing answer may still be wrong
AI-generated content can appear authoritative even when it is inaccurate, incomplete or entirely fabricated.
There have been numerous reported instances of AI systems producing invented facts, legal authorities, quotations and references because the content appeared to fit the prompt they had been given. This is often described as an AI “hallucination”. The wording may be convincing and confidently presented, even though the underlying information does not exist or cannot be verified.
In contract drafting, this may result in provisions being included because they look familiar rather than because they are legally accurate or commercially appropriate. A clause may refer to the wrong legislation, omit an important qualification or conflict with another part of the agreement. Definitions may not align with the provisions in which they are used, while authorities or regulatory requirements cited by the tool may be incorrect or outdated.
The wording may also fail to reflect the agreed deal. An AI-generated contract could include a termination right that undermines the intended contract period or an indemnity that transfers more risk than expected.
These issues can be difficult to identify because the output may read naturally and use recognisable legal terminology. Any factual, legal or regulatory information generated by AI should therefore be independently checked before it is relied upon.
Human review should involve more than correcting grammar and presentation. The reviewer must verify the underlying information, test the contract against the commercial arrangement, consider how its provisions interact and assess what would happen if a dispute arose.
When an AI first draft creates more work
AI is often presented as a way to reduce the time and cost involved in preparing legal documents. However, an inaccurate or poorly structured first draft can sometimes make the legal process more difficult.
A business may use AI to prepare an agreement before asking a solicitor to review it. If the document contains inconsistent clauses, unsuitable legal concepts or provisions copied from a different type of transaction, the solicitor may need to spend additional time identifying and correcting those issues before addressing the commercial arrangement itself.
The same concern can arise in commercial disputes. A party may use AI to prepare correspondence, allegations or a draft defence and then build its position around information that has not been properly verified. Where the factual or legal basis is inaccurate, this can distract from the genuine issues, increase correspondence and add unnecessary time and cost to the dispute.
AI can still be useful when preparing for legal discussions. It may help a business organise information, identify questions, understand unfamiliar terminology or consider the commercial issues it wishes to raise.
However, it should be used to improve preparation rather than replace sound legal advice. An AI-generated starting point is only valuable if its assumptions, facts and legal conclusions are carefully checked.
Confidentiality and intellectual property
The use of public AI tools in contract drafting can create confidentiality risks.
An employee may upload an agreement, heads of terms, pricing structure or negotiation history to obtain a summary or suggested wording. Those documents may contain commercially sensitive information belonging to the business or another party.
The provider’s terms may allow user inputs to be retained, reviewed or used to develop the system. Uploading information could therefore place the business in breach of confidentiality obligations or reduce its control over proprietary information.
Businesses should establish clear rules around the information employees may enter into AI tools. Confidential material, personal data and documents supplied by customers or counterparties should not be uploaded without understanding the safeguards available.
AI-assisted drafting also raises intellectual property questions.
A business should not assume that AI-generated wording is exclusive or proprietary. Similar language may be produced for other users, particularly where the system relies on common contractual patterns.
There may also be uncertainty about rights in content produced largely by an automated system. Businesses should review the provider’s terms to understand the rights granted in relation to inputs and outputs, as well as any warranties or indemnities concerning third-party intellectual property.
Contracts with AI suppliers
Businesses purchasing AI systems should apply the same care to the supplier agreement as they would to any important technology contract.
The agreement should reflect the function the system will perform and the consequences if it fails.
A tool used for low-risk administrative tasks may require relatively straightforward terms. A system used to communicate with customers, process personal data or support important decisions is likely to require stronger protections.
Relevant provisions may include:
AI providers may give limited assurances about the accuracy or suitability of their systems. Some may seek to place responsibility for checking outputs entirely on the customer.
Businesses should understand those limitations before becoming dependent on the service. A broad disclaimer may leave the customer with little practical remedy if an incorrect output causes loss.
Human oversight is part of compliance
Maintaining human involvement is not simply about improving drafting quality. It is also an important part of governance and compliance.
Directors and senior managers remain responsible for the systems used within the business. Responsibility cannot be transferred to an AI tool because it produced the wording or recommended a particular course of action.
Businesses should be able to explain how AI is being used, who approves that use and which decisions require human review.
A reviewer should have enough knowledge and authority to question the output rather than simply accepting it. Human oversight has limited value if the person involved is expected to approve whatever the system produces.
Appropriate records may also be helpful. Where an important contract or decision has been supported by AI, the business may need to show what information was provided, what output was generated and what checks were completed.
Building AI into the wider compliance framework
AI should not be treated solely as an IT issue. Its use may affect legal, financial, operational and reputational risk.
A proportionate framework could begin with a review of the tools already being used, including informal use by employees.
The business can then introduce an AI policy explaining which tools are permitted, what information may be entered and when an output must be reviewed by someone with relevant expertise.
Staff involved in contract drafting or negotiation should receive practical training on confidentiality, data protection, accuracy and the limitations of generated content.
Responsibility for AI governance should be allocated to a senior individual or suitable committee. Higher-risk uses should be assessed before introduction and monitored throughout the period in which the system is used.
Supplier due diligence should also continue after an agreement is signed. AI systems may change through updates, new functionality or alterations to the underlying model. A product that was suitable when first purchased may present different risks as its use expands.
Using AI without losing commercial judgement
AI can support contract preparation and management, but its role should be clearly defined.
It may help produce an initial draft, compare documents, organise information or identify provisions requiring attention. It can also help a business improve its knowledge before entering negotiations or seeking legal advice.
However, an AI-generated draft is not necessarily a reliable shortcut. Where its content is inaccurate, inconsistent or based on the wrong assumptions, correcting it may create more work and cost than preparing the document properly from the outset.
AI cannot determine the commercial outcome a business should pursue, verify every factual or legal assertion or decide whether a particular allocation of risk is acceptable. Those decisions require an understanding of the transaction, the legal framework and the organisation’s wider objectives.
Businesses should therefore use AI to support their commercial position rather than allowing it to replace professional judgement. The most effective approach is likely to combine technology with informed legal advice, careful fact-checking and appropriate human oversight.
stevensdrake’s Corporate and Commercial team can advise businesses on commercial contracts, agreements with AI and technology providers, internal AI policies and wider governance and compliance arrangements. Early advice can help businesses benefit from new technology without overlooking the contractual protections and human judgement they continue to need.
Nusrat Qureishi joins us as the Head of Corporate & Commercial Law. She is a corporate commercial lawyer with more than 20 years’ experience with particular expertise in the sale and purchase of businesses and companies, as well as the establishment of partnerships and other joint ventures.
Having trained and worked for many years for a City Law firm, Nusrat has a wealth of experience of working on high-value and complex corporate transactions. Her experience extends to advising on mergers, investment, shareholder, restructuring, banking and re-financing and corporate governance work. Her clients have included entrepreneurs, owner-managers, mid-market companies, SMEs, PLCs, international corporates, private equity, banks and financial institutions, joint venture parties, LLPs, partnerships and individuals.
Client service is highly valued by Nusrat who invests time to gain a real understanding of her clients’ needs to meet their goals as effectively and commercially as possible. She aims to be involved in the preparation of a business for sale or purchase from an early stage to ensure that the process runs smoothly and without unnecessary risk, helping to reduce both time and cost for her clients. Her experience allows her to advise companies on strategy – enabling entities to realise the next stage of their growth plans and also support more sophisticated businesses to achieve their commercial objectives.
Nusrat’s approach is totally consistent with the ethos of stevensdrake where client satisfaction is valued together with the provision of a high quality service.
stevensdrake © 2019 All Rights Reserved. Authorised and regulated by the Solicitors Regulation Authority Number 525527. *stevensdrake* and *stevensdrake solicitors* are trading names of stevensdrake limited (registered in England and Wales under number 7159192). A list of the directors is displayed at the above address. stevensdrake are authorised and regulated by the Financial Conduct Authority (FCA) for debt collection and insurance mediation. It is registered the FCA under registration number 702596.
