AI License Renewal: A 90-Day Negotiation Checklist

An AI license renewal is the moment when a company can reduce seats, change tools or renegotiate terms without penalties. Prepare it 90 days before expiry with a checklist: usage data, seats to cut, overlaps, clauses to verify, alternatives and a calendar. Whoever arrives at the negotiation with the numbers negotiates from a different position than whoever receives the renewal quote.

Why 90 days before expiry?

Ninety days is a practical horizon, not a rule. It gives you time to collect usage data for a meaningful period, consult the departments, have legal and procurement assess the contract and, if needed, test an alternative. With less time, renewal happens by inertia.

Two constraints can bring deadlines forward. The first is the notice period for cancellation or for changing volumes, which many contracts set well before the renewal date. The second is automatic renewal: if you do not act within the stated window, the commitment repeats on the previous terms. The first action is therefore to read the contract and note the dates, then confirm them with legal and procurement so that nobody works from an outdated deadline.

The preparation calendar

An example with renewal on 31 January 2027, so with preparation starting on 2 November 2026:

• T-90 (2 November 2026): kickoff. Identify the owner, contracts and notice dates, and extract the data.
• T-60 (2 December 2026): analysis closed. You have seats to cut, overlaps and scenarios. Consult the departments.
• T-30 (1 January 2027): negotiating position defined, requests sent to the vendor, internal decision on term and volumes.
• Expiry (31 January 2027): signature or exit, with the new configuration already planned.

If the contract requires 60 days' notice, the calendar must be brought forward accordingly: the reference date is always the nearer of renewal and cancellation deadline.

Checklist: the data to collect

The first block is quantitative. For each tool, prepare:

1. Seats purchased, assigned and active, with the definition of activity used. The method is in how to find unused seats.
2. Cost per seat and cost per active user, as explained in real cost per AI seat.
3. Usage by department and role, in aggregate form, to know where value is concentrated.
4. Use cases covered, from the use case register or from interviews with department heads.
5. Ancillary costs: training, administration, integrations, following a total cost of ownership approach.
6. Adoption trend over the last year: growth, stability or decline.

Without these numbers, you cannot challenge the vendor's proposal. The vendor sets the scope of the renewal using its own data, and usually starts from current volume. In the renewal quote, moreover, seats appear as a single total: the distinction between assigned and active exists only in your data.

A practical tip: collect data on all tools right away, including those that expire later. The overall view lets you shift volumes from one tool to another, instead of treating each renewal as an isolated file.

Checklist: seats to cut and overlaps

The second block turns data into a concrete request.

• Inactive and sporadic seats: list by department, already validated with department heads and cleaned of false positives.
• Seats to keep with reservations: licenses for seasonal roles or in activation, to be reviewed after a defined period.
• Expected growth: hires or new projects requiring additional seats, to be forecast with a realistic margin, not an inflated one.
• Overlaps: use cases covered by several tools for the same people. The detection method is covered in how to detect overlap between AI tools.
• Volume scenarios: at least three hypotheses (current volumes, reduced volumes, reduced volumes with consolidation), each with its expected cost.

An example of the approach, without prices: if out of 600 seats 450 are active and another 30 are sporadic, the reduced scenario starts from 450 seats plus a justified growth margin, not from 600.

Checklist: clauses to verify, in general terms

Clauses decide how much flexibility remains after signing. Read them with legal and procurement. In general terms, verify:

• Flexibility on volumes: whether seats can be reduced during the term or only at renewal, whether there is a binding minimum, and how increases work.
• Term and renewal: length of the commitment, automatic renewal, notice to cancel or modify.
• Exit: termination conditions, any exit costs, service shutdown times.
• Data: what happens to data and content at the end of the relationship, how long they remain available and in what format they can be exported. The topic also involves the DPO, because it touches the processing of personal data under the GDPR (Article 28 on processors).
• Changes to terms: how and with what notice the vendor can change features, plans or service terms.
• Service levels and support: what is guaranteed and what remedies are provided.

This list is a reading guide, not legal advice: the terms of each contract must be verified against the signed text.

Checklist: alternatives and negotiating leverage

The strongest leverage is a credible alternative. You do not need to have implemented it: you need to have assessed it.

• Identify one or two alternative tools for the highest-spend use cases.
• Estimate migration and training cost, to know how much switching is really worth. The framework is in when consolidating AI tools pays off.
• Assess the option of a shorter term or partial volumes on trial.
• Decide in advance your minimum acceptable outcome, and what you would do if you did not get it.

Compare tools by use case, not by the preferences of individual teams. A comparison built on shared criteria is easier to defend with finance and with the vendor.

How to run the negotiation

Negotiation is not limited to price. It covers volumes, term, flexibility and exit conditions, and it goes better when requests are few, backed by data and already agreed internally. A list of ten requests of equal weight dilutes your position: distinguish what is indispensable from what is desirable.

Some rules of method, valid regardless of vendor:

1. A single internal counterpart, who coordinates IT, procurement, finance and departments.
2. Data before price. Open with usage and use cases, then move to terms.
3. Written, comparable requests, so you can compare offers over time.
4. No decisions under deadline pressure. If needed, ask for a short extension to complete the assessment.
5. Document the outcome and the reasons, for the next renewal.

After signing, schedule a check at three months to verify that the new configuration holds and that adoption has not dropped. Also note the rejected requests and the reasons: they will be the starting point of the next cycle, together with updated usage data.

Where Delphis fits

Delphis is a vendor-independent platform that connects spend, adoption, impact and regulatory obligations. It prepares the data the negotiation needs: contracts and renewal dates, purchased versus active seats, usage by department, overlaps and savings simulation by scenario. Impact is measured against a baseline, and data on people stays aggregated. More details are on the Spend page.