Lease Heads of Terms Generator

Lease heads of terms (HOT) set the commercial skeleton of a deal before solicitors turn it into a full lease. For a leasing negotiator, the first draft is where time is lost: rent, term, breaks, incentives, and repair obligations must be written clearly, and they must sit inside the landlord's approved position rather than inventing a one-off bargain.

This use case uses an LLM to assemble a first-pass HOT from structured deal parameters, checked against the landlord's standard position. The output is a negotiator-ready draft, not a sent document. The model proposes wording and flags where the proposed deal sits relative to the standard; the negotiator still reviews, edits, and issues the terms to the other side.

The aim is speed with guardrails. A compliant first-pass in minutes replaces a blank page and a memory hunt through prior deals, while keeping authority with the human who owns the relationship and the paper trail.

The draft only runs when both sides of the input set are present.

Deal parameters are the live commercial facts for this unit and this counterparty. Typical fields include:

• Rent (base rent, review basis, and any stepped or indexed structure)
• Term length and commencement assumptions
• Break rights (who can break, when, and on what notice or conditions)
• Incentives (rent-free, capital contribution, fit-out support, or other concessions)
• Use, alienation, repair, and service-charge framing if those are part of the HOT pack for this asset class
• Any agreed deviations already logged in the deal file

The landlord's standard position is the approved baseline for this portfolio or asset type: preferred wording, acceptable ranges, hard no-go items, and the usual trade-offs for incentives versus term or for breaks versus rent. It may live as a playbook, a clause bank, prior approved HOT templates, or a structured policy table. Whatever the form, it must be available to the model as reference material for this run.

If deal parameters are incomplete, or the standard position cannot be retrieved for the relevant landlord, asset, or lease product, the system returns empty output and does not invent missing commercial terms. Silence is safer than a plausible but unauthorised draft. The negotiator then completes the inputs or escalates missing policy coverage before retrying.

related

Once inputs are complete, the model maps each deal parameter onto the standard position and drafts HOT sections in the landlord's usual order and tone.

Where the proposed deal matches the standard, the draft uses approved phrasing and ranges with little commentary. Where it stretches the standard (for example a longer rent-free period, an earlier tenant break, or a softer repair package), the draft still produces workable wording but marks the stretch so the negotiator can see it before anything leaves the desk. The marking can be inline notes, a short variance list, or both, depending on how the team reviews work.

The model does not invent market "norms" or fabricate peer deals. It works from the parameters and the standard you supplied. Ambiguous fields stay unresolved rather than being guessed. If a parameter conflicts with a hard rule in the standard (for example a break timing the policy forbids), the draft either withholds that section or surfaces the conflict explicitly, per your operating rule, instead of quietly softening the position.

The result is a coherent first-pass HOT: commercial points filled, structure familiar to your solicitors and counterparts, and variances visible. That is the speed gain: minutes to a reviewable draft instead of a slow rebuild from prior emails and templates.

related

Human-in-the-loop is non-negotiable. The LLM drafts; the negotiator still sends.

Before issue, the negotiator confirms that rent, term, breaks, and incentives match the live deal, that variances from the standard are intentional or escalated, and that counterpart names, unit descriptions, and dates are correct. They also decide tone and strategy: how firm to sound on a stretch point, whether to hold an incentive for later rounds, and whether legal should see the pack before it goes out.

The generator does not replace authority matrices, board approvals, or solicitor review of complex structures. It does not negotiate. It does not accept counter-HOT wording without a fresh human pass. Audit trails should record who approved the draft, which standard version was used, and which parameters fed the run, so later disputes about "what we offered" have a clear source.

Empty output on missing inputs is part of that control. Automating past a gap would produce terms nobody authorised. Better to stop, fix the input, and regenerate.

HOT drafting sits mid-funnel. Screening and scoring decide whether a tenant is worth terms at all. Dynamic rent pricing at lease-up informs what rent and incentive envelope the deal parameters should carry. Compliance checking later (or in parallel) tests whether the emerging lease text still matches policy and statute once solicitors expand the HOT into full form.

Used together, those steps reduce rework: you do not draft generous terms for a weak credit path, you do not lock rent language that contradicts the pricing decision, and you do not discover standard-position breaches only after legal has spent hours on the long form. The related pages on AI Tenant Screening & Scoring, Dynamic Rent Pricing at Lease-Up, and Lease Compliance Checker cover those adjacent controls in more depth.

This page's scope stays narrow: generate a compliant first-pass HOT from parameters plus standard position, hand it to the negotiator, and refuse to draft when either input set is missing.

Keep the standard position versioned and scoped (by landlord, asset class, or lease product) so the model does not pull the wrong baseline. Keep deal parameters in one place for the unit and counterparty so rent, break, and incentive fields do not drift between CRM notes and the generator. Train reviewers to treat variance flags as decision prompts, not decoration.

Measure success in cycle time to first issued HOT and in how often issued drafts need commercial rewrite for standard-position mistakes, not in how often the model sounds fluent. Fluent empty guesses are failures; empty output on incomplete inputs is correct behaviour.

When those habits are in place, leasing negotiators spend less time assembling the skeleton and more time on the judgement calls that actually close the deal: which stretch to allow, which to trade, and when to walk away.