How to build better contracting context

AI can process a contract in seconds. That doesn’t necessarily mean it understands how your organization wants that contract handled.

A model might know what an indemnity clause is, for example, but it won’t automatically know your preferred position, whether your business is willing to compromise or which deviations need senior approval.

That’s business context – and it’s becoming more important to get the most value out of AI.

Summize’s recent AI Fluency Report found that 75% of in-house legal professionals say their AI tools always or frequently reflect their organization’s policies, preferences or historical decisions. But at the same time, only 15% have AI connected directly to approved internal knowledge sources. The other 85%? They’re manually applying context every time.

That suggests the challenge isn’t whether Legal understands that context matters – it’s how that context gets captured, structured and made available to AI in a way that’s consistent and useful for the entire business.

What does business context mean?

Context is the information AI needs to understand how your organization actually works – from its preferred positions and risk appetite to approval processes and business priorities. Our Strategic Implementation Director Lara Trope explains that context exists on several different levels:

These layers build on each other – a lawyer might know that a particular clause is unusual, the team might have a documented fallback position and the organization might have a commercial reason for accepting a particular level of risk. When AI can draw on all three, it has a more complete picture of the situation, helping it produce outputs that are more relevant, consistent and aligned with how your organization actually works.

Why does AI need business context?

• More relevant outputs – AI can take your organization’s policies, preferences and positions into account rather than generating a generic response.

• Greater consistency – shared context can help different lawyers reach more consistent conclusions when reviewing similar contracts.

• Less repetitive prompting – the more useful knowledge available to the system, the less lawyers need to explain the same amends manually every time.

• Better foundations for automation – when organizational knowledge is structured and reusable, it becomes easier to support repeatable workflows rather than simply repeating individual AI interactions.

“Without this context, AI can still produce an answer that looks reasonable and correct, but it might not be the answer that makes sense for your business. For example, you might generally reject uncapped liability, but accept it for a strategic customer where the commercial value justifies that risk. Capturing that reasoning is more useful than simply recording ‘reject uncapped liability'."
‍– Lara Trope, Strategic Implementation Director

Lara Trope’s four easy ways to start building context

Building contracting context can sound like a major knowledge-management exercise. But it doesn’t have to be.

“You don’t need to capture every historical decision, document every possible scenario or create a perfect playbook before anyone can use AI. Start small, use real work to test what you have and build from there.”
‍– Lara Trope, Strategic Implementation Director

1. Start with one high-volume contract type

The easiest place to begin is usually somewhere your team already has a lot of experience.

Rather than trying to operationalize every contract at once, choose a contract type that comes through regularly, like an NDA, MSA or a particular type of commercial agreement. To make this easier, ask:

• Where do we see the same issues repeatedly?
• Which clauses generate the most questions or negotiations?
• Where are lawyers making similar decisions again and again?
• Which contract type would benefit most from greater consistency?

Starting with one area gives you something manageable to test. It also makes it easier to see whether the context you capture is actually improving the work.

2. Capture what you and your legal team already know

You probably already have a lot of the context you need without even realizing it. It just might be living in your team’s heads, old negotiations, email threads or informal conversations. This is the time to get it somewhere the wider team can reuse it.

For the contract type, ask:

• What do we accept? Which positions are generally fine without further discussion? ‍
• What do we reject? Which terms or positions fall outside the organization’s risk appetite? ‍
• Where is there flexibility? Which positions can change depending on the deal, counterparty or commercial circumstances? ‍
• What needs escalation? When should a lawyer involve a senior colleague, another function or the business?

You don’t need a sophisticated system to do this initially. A shared document can be enough to start turning individual expertise into something the team can access and improve together.

3. Turn recurring decisions into reusable knowledge

Once you’ve started capturing those decisions, look for patterns. If lawyers are repeatedly giving the same advice, asking the same questions or making the same judgment calls, that knowledge may be worth turning into something reusable.

Depending on the use case, that could mean…

• Contract playbooks
• Standard templates
• Fallback positions
• Approval rules
• Review guidance
• Examples of acceptable and unacceptable language
• Escalation criteria
• Saved prompts

This is where context starts moving from individual knowledge to team knowledge.

Instead of one lawyer knowing how the organization approaches a particular clause, the position becomes something others can access and apply.

And because these resources can be updated as the business changes, they don’t have to be treated as permanent documents. They can become living sources of organizational knowledge.

4. Test your context with a real AI use case

Don’t spend months building a context layer without testing whether it actually helps. Take one real use case and give the AI the relevant organizational context. Then see what happens.

For example, you could use a small set of approved positions and review guidance to test how effectively AI can identify deviations from your preferred position in a particular contract type.

Pay attention to where the output is useful and where it falls short. Here are a few questions to ask:

• Does the AI have enough information to understand the organization’s position?
• Is the guidance clear enough?
• Are there exceptions that haven’t been captured?
• Does the output reflect the business’s actual risk appetite?
• Are there decisions that still need human judgment?

Those gaps are useful. They tell you what context is missing and where your knowledge needs to become more explicit.

Lara’s biggest lesson: you don’t need a perfect context layer to get started

One of the biggest mistakes is treating organizational context as something that needs to be completely documented before it can be useful, but that’s not the case.

Your legal team already has a huge amount of knowledge about how the business contracts, negotiates and manages risk. The first step is simply to make some of that knowledge visible and reusable.

Start with one contract type. Capture the decisions that come up most often. Turn useful patterns into playbooks, guidance or other reusable knowledge. Then test that context against real legal work.

Over time, those small pieces can build into a much more connected picture of how your organization operates.

For more insights into AI in legal, building out your business context and the future of AI in the in-house world, check out our full AI Fluency Report below.