Counter-position generator

When counsel reviews a vendor's markup, the slow part is rarely spotting the problem. It is deciding what to send back: which fallback language fits the gap, how far the playbook allows you to move, and in what order to propose alternatives if the counterparty pushes again. A counter-position generator automates that decision layer. For each disputed clause, it proposes ranked fallback counter-redlines tied to playbook acceptance thresholds, so reviewers start from defensible positions instead of blank redlines.

The generator does not replace counsel. It prepares structured counter-proposals that map directly to approved playbook language. If the playbook has no approved fallback for a deviation, the output for that clause is empty. That silence is intentional: it flags gaps in fallback coverage rather than inventing language on the fly.

For every flagged deviation, the generator returns one or more counter-redlines ordered by preference. Each entry is a concrete edit proposal: suggested replacement or modified text, scoped to the exact span in the counterparty's draft.

Every counter includes three grounding fields:

• Clause span: the character or paragraph range the counter applies to, so reviewers and downstream tools can align edits without manual re-markup.
• Playbook tier: the acceptance band the counter satisfies (for example, Tier 1 preferred, Tier 2 acceptable with conditions, Tier 3 escalation-only).
• Fallback clause ID: the stable identifier for the approved fallback language in the playbook library.

If no fallback clause ID applies at any tier, the generator returns nothing for that dispute. Counsel still reviews, edits, and sends redlines. The generator accelerates drafting; it does not auto-send markup or bind the organization to proposed language.

Playbooks encode more than "yes" or "no." They define graduated tolerance: preferred positions, acceptable compromises, and positions that require business or executive sign-off. The counter-position generator reads those thresholds and ranks outputs accordingly.

A typical ranking logic:

1. First choice: the highest-tier fallback that fully resolves the deviation without crossing a hard limit (uncapped liability, unrestricted IP assignment, missing audit rights).
2. Second choice: a Tier 2 fallback that trades a controlled concession for movement on the counterparty's ask ( narrower cap, longer notice period, mutual instead of one-way obligation).
3. Third choice: a Tier 3 or conditional fallback flagged for escalation, included only when the playbook explicitly permits proposing it before approval.

Ranking is deterministic given the same playbook version and clause input. That repeatability matters for audit trails and for training junior reviewers: the same deviation should surface the same ordered options unless the playbook changes.

The generator does not score "likelihood the vendor accepts." That prediction belongs elsewhere in the negotiate stack. Here, the quality bar is fidelity to approved fallback language and correct tier assignment.

Outcome for this use case is quality: every counter must be traceable, tier-correct, and empty when the playbook offers no approved path.

Traceability. Each counter-redline links a specific clause span to a specific fallback clause ID. Reviewers can open the source fallback in the library and confirm the proposed text matches approved wording, not a paraphrase that drifted in generation.

Tier correctness. Mislabeled tiers are a common failure mode. A Tier 2 compromise presented as Tier 1 creates compliance risk; an escalation-only position offered as routine slows deals incorrectly. Quality review should spot-check tier labels against the playbook's acceptance matrix for high-risk clauses (indemnity, limitation of liability, data processing, termination).

Empty when appropriate. An empty result on a disputed clause is a signal, not a bug. It usually means fallback clause generation has not yet produced an approved alternative for that deviation pattern, or the playbook deviation report flagged something outside current coverage. Counsel drafts manually or triggers playbook expansion before the next negotiation cycle.

No synthetic language. The generator proposes from the fallback library. It should not fabricate novel contractual terms to fill gaps. That constraint preserves enforceability and keeps legal sign-off meaningful.

The counter-position generator typically runs after deviation detection and before redlines leave legal review.

1. Deviation identification. Incoming markup is compared to the playbook. Disputed clauses enter a work queue with deviation type and severity.
2. Counter-position generation. For each dispute, ranked fallback counter-redlines are proposed with span, tier, and fallback clause ID.
3. Acceptance gating. Proposed counters can flow through an auto-redline acceptance gate that auto-approves Tier 1 matches and routes Tier 2+ for human review.
4. Impact assessment. When a counter involves a concession, a concession impact predictor can estimate commercial or risk exposure before counsel commits to that fallback in the outbound redline.
5. Counsel send. Reviewers edit, combine, or override proposals, then send redlines through normal channels.

This sequencing keeps generation focused on playbook fidelity while gating and impact tools handle approval policy and business trade-offs.

Inputs usually include: the marked-up agreement (or extracted clause objects), deviation metadata from the playbook comparison layer, the active playbook version with tier definitions and fallback clause library, and optional business context (deal size, data sensitivity) that some playbooks use to select among tier-allowed fallbacks.

Outputs are structured records per disputed clause: ordered list of counter-redlines, each with clause span, playbook tier, fallback clause ID, and the proposed replacement text pulled from the library. Aggregated views can show coverage gaps (disputes with zero counters) for playbook maintenance.

Integration with contract lifecycle and AI legal platforms varies by vendor, but the pattern is consistent: ingest markup, call generation against playbook APIs or embedded libraries, return proposals into the redline or review UI.

| Vendor | Typical role | |--------|----------------| | Ironclad | Playbook and workflow host; counters often surface inside Ironclad's review flows tied to clause types and approval paths. | | LegalOn | Playbook-guided review; generation aligns with LegalOn's fallback and guidance models for Word-native markup. | | Litera | Negotiate and compare tooling; ranked counters can feed Litera's redline assembly and version comparison. | | Harvey | AI-assisted legal drafting; Harvey may draft or refine counter language, but tier and fallback ID should still anchor to the authoritative playbook library for quality control. |

Regardless of vendor, treat the playbook library as source of truth for fallback text and IDs. AI layers propose and rank; they should not become a shadow playbook.

Keep fallbacks current. The generator is only as good as fallback clause generation and playbook curation. When empty results cluster on the same deviation type, expand approved fallbacks rather than asking reviewers to improvise each time.

Use deviation reports for coverage. Regular playbook deviation reports show which counterparty patterns lack fallbacks. Prioritize library work there before tuning ranking logic.

Preserve counsel judgment. Ranked counters are starting points. Counsel merges clauses, adjusts tone for relationship context, and may reject a Tier 1 proposal when deal-specific facts warrant escalation.

Version the playbook. Counter outputs should record playbook version alongside fallback clause ID so later audits can reconstruct why a given redline was proposed.

Measure quality, not volume. Success is measured by traceable, tier-correct proposals and appropriate empties, not by maximizing the number of counters generated per document.

A counter-position generator turns playbook thresholds into actionable redline drafts. Ranked fallback counter-redlines give negotiators defensible first responses on every covered dispute, clear gaps where the playbook does not, and a consistent audit trail from clause span to approved fallback. Counsel still sends the redlines; the generator ensures those redlines start from the organization's actual negotiating positions, not from memory or ad hoc search.

[REDACTED]