The Contract Manager's Day With AI — Where It Helps, and Where It Breaks

Our AI Readiness Benchmark scored the Contract Manager role and gave you a number. But a number doesn't tell you what happens when an AI tool runs into a supplier who won't play ball, or a liability clause that doesn't quite fit the playbook, or a SLA dispute where both sides have a reasonable argument.

This piece does that. We've built a hypothetical scenario — realistic enough that most contract managers will recognize the situations — and walked through the full contract lifecycle to show exactly where AI accelerates the work and exactly where it trips.

Meet Sarah, Contract Manager at a mid-sized manufacturing company. She's just received a sourcing handoff from the procurement team. They've concluded a competitive RFx for a critical IT services contract — a three-year managed services deal with a supplier called Nexura. Commercial terms are agreed. Now it's Sarah's job to get this contracted.

Drafting the first version of the contract is where AI genuinely earns its keep. Sarah's organization has a standard IT services template and a contract playbook with pre-approved positions on liability, IP, data protection, and confidentiality. She feeds the sourcing handoff — agreed pricing, volumes, SLAs, term, renewal and termination mechanics — into an AI-assisted CLM tool.

The AI populates the template accurately. It cross-references the handoff pack against the draft and flags where agreed terms haven't been reflected. It applies playbook starting positions to the risk clauses without being asked. It produces a clean first draft in less than an hour that would have taken Sarah the better part of a day to build manually.

This is AI doing exactly what it's built for: structured template work with clear inputs, clear rules, and a clear definition of what a good output looks like.

Extracting and logging obligations once the contract is executed is another win. The AI reads the final contract, extracts all dated obligations — delivery milestones, SLA thresholds, payment terms, notice periods, certification renewal dates — and populates the tracking system. It sets alerts at 90 and 30 days before expiry, flags the auto-renewal clause with the notice window, and links the contract record to the sourcing event and supplier master. Things Sarah would otherwise do manually, checking fields against a checklist, are done in minutes.

Identifying non-standard clauses during the redline process is also where AI adds real value. When Nexura comes back with their redlined version, the AI runs a comparison against the playbook and immediately surfaces three deviations: a liability cap set at the total contract value rather than twice contract value, a proposed exclusion of consequential losses with no carve-out for data breach, and an IP ownership clause that assigns jointly developed work to Nexura. It flags all three for Sarah's attention with a brief summary of why each deviates from the standard position. What might have taken Sarah an hour of careful clause-by-clause review is in front of her in seconds.

Here's where it gets interesting.

Trip #1: Reading the room in negotiation

Sarah reviews the AI's redline summary and agrees on the positions she wants to hold. She asks the AI to help her think through where to apply the playbook fallbacks — the pre-approved concession positions — and where to hold firm.

The AI applies the playbook correctly. It knows the fallback on liability is to accept twice contract value if the supplier won't move to three times. It knows the fallback on consequential loss is to accept the exclusion if data breach and IP infringement carve-outs are preserved.

What it doesn't know is that Nexura's negotiator called Sarah's colleague last week to say they're under commercial pressure to close this quarter. That the supplier's procurement team has already flagged this contract internally as a priority win. That Nexura needs this deal more than they're letting on.

Sarah knows all of this — from context, from relationship history, from reading between the lines of the call. She decides to hold on the liability cap and not deploy the fallback yet. She's reading leverage. The AI can't do that. It would have moved to fallback positions on schedule, leaving value on the table.

Trip #2: The qualitative SLA dispute

The contract is executed. Three months in, Nexura is hitting their uptime SLA — 99.5% availability, measured and confirmed. But Sarah's operations team is frustrated. The system is technically up, but response times for support tickets are consistently slow, escalations aren't being managed proactively, and the "strategic guidance" the contract promised from Nexura's senior team hasn't materialised in any meaningful way.

Sarah asks the AI to assess Nexura's SLA performance. The AI reviews the data, checks it against the contractual definitions, and reports back: all quantitative SLAs are met. Credit deductions: nil.

But the operations team is right to be frustrated. The contract includes a clause requiring Nexura to provide "commercially reasonable efforts to support the buyer's operational objectives" alongside the quantitative SLAs. The question of whether Nexura is meeting that obligation isn't a data question. It requires a human to define, in context, what commercially reasonable looks like for this engagement — and then to make a judgment call about whether Nexura's behaviour crosses the line from disappointing to contractually deficient.

The AI reads the literal text. It cannot assess whether Nexura is meeting the spirit of the agreement. Sarah has to make that call — and if she decides Nexura isn't meeting the standard, she then has to issue a formal notice that is carefully worded to preserve her legal position without unnecessarily damaging the relationship. That's a judgment and a relationship call the AI can't own.

Trip #3: The liability clause Nexura won't move on

Negotiations stall. Nexura is holding firm on excluding consequential losses without a data breach carve-out. They claim their legal team won't budge.

The AI's input: the current playbook position requires the carve-out, the fallback is to accept the exclusion if IP infringement carve-out is retained, the red line is a blanket exclusion with no carve-outs at all. Nexura's proposal is between the fallback and the red line — it preserves IP indemnification but not data breach.

The AI can tell Sarah where Nexura's proposal sits on the playbook spectrum. It can't tell her whether the data breach exposure is likely to materialise given Nexura's role in the environment, whether Nexura's legal team is genuinely constrained or posturing, or whether Sarah's General Counsel — who Sarah knows has flagged data breach exposure as a current priority concern internally — would want to be escalated to directly rather than going through the standard playbook process.

Sarah escalates. The AI didn't know she should.

Trip #4: The handoff

The contract is executed and Sarah needs to brief Marcus, the Supplier Relationship Manager who will own the ongoing relationship with Nexura.

The AI produces a contract summary — key terms, SLA thresholds, obligation milestones, non-standard clauses, the data breach carve-out issue and how it was resolved. It's accurate and complete as a document.

What the document can't capture is that Nexura's commercial lead was prickly throughout negotiation, pushed hard and unsuccessfully on payment terms, and made a comment in the final call that suggested they'd be looking to find scope arguments during the contract to recover the margin they conceded. Marcus needs to know that. Not because it's in the contract, but because it will shape how he manages the first twelve months of the relationship.

Sarah tells Marcus this in a conversation. The AI wrote the briefing pack. The intelligence that makes the briefing pack useful came from Sarah's memory of how the negotiation actually felt.

The AI Readiness Benchmark scored the Contract Manager role using a weighted methodology — tasks are scored by how ready AI is to perform them, weighted by the human effort those tasks actually consume. The Contract Manager score reflects a role where AI can genuinely automate significant portions of the structured work: drafting, clause extraction, obligation tracking, alert configuration, status reporting.

But the high-effort tasks — contract negotiation, risk allocation, SLA interpretation, dispute management — are where the weighting matters most. These are the tasks that consume the most time, carry the most risk, and require the kinds of judgment, relationship reading, and organizational context that AI cannot yet replicate.

The practical implication for procurement leaders isn't "don't use AI for contract management." It's more specific than that: know which tasks you're deploying AI on, and make sure a human is still owning the inflection points where structured rules meet messy reality.

Sarah's story ends well. The contract gets signed. Nexura performs adequately, and the SLA conversation gets resolved with a formal letter and an agreed interpretation. But the moments that determined those outcomes weren't the ones the AI handled.

The AI Readiness Benchmark for procurement roles is available at air.procurement.news. The benchmark scores five Sourcing & Contracting roles using a weighted methodology based on task-level AI readiness assessment. Research conducted in partnership with AllCaps.ai .