Gap Analysis — Concordia vs. Pactum and Nibble

Analysis date: 2026-04-23

Scope: Phase 179 Concordia task 179.1.1.4.

This analysis measures Phase 179 Concordia against Pactum and Nibble, the two most mature autonomous-negotiation commercial systems in the audit. It complements the Mediator.ai gap analysis at gap-analysis-mediator-ai.md. Where Mediator.ai defines the consumer cooperative-negotiation frontier, Pactum and Nibble define the enterprise procurement and commerce-negotiation frontier — the category Concordia's Maat (§179.7.1), Aglaea / Freya (§179.7.5), and agent-to-agent (§179.7.6) integrations must surpass.

The seven gap axes required by 179.1.1.4:

1. Procurement and commerce negotiation workflows
2. ERP and marketplace integration
3. Supplier audit trails
4. Objective savings metrics
5. Supplier satisfaction metrics
6. Campaign management
7. Guardrailed autonomous execution

For each axis: the Pactum baseline, the Nibble baseline, the Concordia requirement, the gap, the Phase 179 task pointers that close it, and an evidence note. The audit source material lives in mediation-negotiation-2026.md §8 and §9; the tagged sources are in source-matrix.md §B.

0. Baseline summaries#

Pactum#

• Category: Enterprise procurement negotiation (agentic AI for supplier negotiations). Global 2000, Fortune 500 buyers.
• Autonomy: Variable and configurable. "Autonomously or with buyer approval" within guardrails set by procurement policy. Highest effective autonomy in the audit cohort.
• Architecture (public): AI agents embedded in enterprise systems via standard APIs; integrate with ERP / procurement / CMS; chat-style negotiation with suppliers; policy-bounded decisioning. Specific LLMs and algorithms not disclosed.
• Compliance: SOC 2 Type II certified. Full negotiation audit trails. Trust portal at trust.pactum.com.
• Deployment: Claims 2–4 week deployment. Named customers include Walmart, Honeywell, Bristol Myers Squibb, Veritiv, Suez, Linde, Novartis, Tetra Pak, Mediclinic, Otto, Global Industrial, Vallen.
• Outcomes (third-party verified for Walmart, Sourcing Journal): 3% average savings, +35 days payment terms. 60+ Global 2000 enterprises. $54M Series C in June 2025 led by Insight Partners. 489% YoY spend growth handled, 2.5× ARR.
• Scope: Bilateral negotiations run massively in parallel. Not multi-party. Category-specific to procurement spend (price, payment terms, rebates, service levels).

Nibble#

• Category: Commerce- and procurement-side negotiation chatbot. Originally DTC e-commerce make-an-offer; expanded into B2B procurement.
• Autonomy: High. Handles mass renegotiations autonomously via self-serve portal or API; negotiates with hundreds of suppliers or thousands of shoppers in parallel.
• Architecture (public): Agentic AI with custom LLM guardrails where "pricing functions" are kept "secure, controlled and never seen by any LLM" — the LLM handles language, a deterministic pricing engine holds the numbers.
• Compliance: ISO 27001 certified. "Your data is your data"; data not used to train other implementations.
• Deployment: Plugins for Shopify, Magento, Adobe Commerce, Coupa, SAP Ariba, and CRMs.
• Outcomes (self-reported): 350+ organizations; ~30,000 negotiations per month; 2M+ cumulative automated negotiations; deal range $5 to six figures.
• Scope: Bilateral per conversation, massive parallelism. Buy-side and sell-side.

1. Procurement and commerce negotiation workflows#

Pactum baseline. Deep bilateral procurement workflows for supplier pricing, payment terms, rebates, service levels, joint business plans, capital allocation, shared services, hiring offers, board resolutions, and cross-company resource conflicts (per their procurement-agents page). Pre-sourcing and post-sourcing flows are first-class; Pactum specifically targets "mid-tier and tail spend" where bilateral human negotiation is uneconomic.

Nibble baseline. DTC make-an-offer flows (shopper → merchant), mass supplier renegotiation (merchant → many suppliers at once), post-sale discount / buy-back flows. Focused on price and payment terms; contract harmonization on the procurement side.

Concordia requirement. Integrate with @maat/strategy, @maat/finance, @maat/supply-chain, and @maat/agents for supplier pricing, payment terms, rebates, service levels, joint business plans, capital allocation, shared services, hiring offers, board resolutions, cross-company resource conflicts (179.7.1.1). Add working-capital optimization: DPO extension, early payment discounts, volume commitments, quality guarantees, warranty terms, supplier relationship scoring (179.7.1.3). Add Pactum-class procurement agents for Asase, Freya, Cybele, Brigid, Saraswati, Aje, and Maat portfolio companies (179.7.1.2). Add returns, custom-order disputes, fit/alteration compromises, pricing offers, resale authenticity, creator collabs, and luxury-service recovery (179.7.5.1).

The gap. Concordia must match Pactum on the procurement bilateral workflows and surpass it on (a) multi-party supplier cohort negotiations (where Pactum runs parallel bilaterals, Concordia can model coalition stability and Shapley attribution for joint volume commitments), (b) non-price terms that Pactum does not treat as first-class (audit rights, data-sharing, sustainability clauses, originality / IP provenance when applicable), and (c) cross-domain workflows — Concordia's procurement work inherits the same bargaining substrate used in Themis governance and Kuanyin restorative flows.

How Concordia closes it.

• 179.7.1.1–179.7.1.5 — full Maat / portfolio procurement map.
• 179.4.2.4 — CP-SAT / MILP for discrete procurement constraints (schedules, budgets, ownership percentages, milestones, roster assignments, routing, shipment terms, resource allocation).
• 179.4.2.7 — coalition stability for supplier-cohort deals.
• 179.2.3.2 — agreement DSL covers money, equity, royalties, deadlines, deliverables, service levels, licensing, exclusivity, territory, data rights, warranty, confidentiality — the full procurement term surface.

Evidence note. Pactum's Walmart-reported 3% savings + 35-day DPO extension is the strongest verifiable procurement baseline in the audit (Sourcing Journal, strength B). Concordia's procurement agents have a concrete outcome target to beat in 179.8.2.3.

2. ERP and marketplace integration#

Pactum baseline. "Embedded in enterprise systems via standard APIs," integrates with ERP / procurement / CMS. 2–4 week deployment. Specific integrations not publicly enumerated beyond the enterprise-grade posture.

Nibble baseline. Explicit plugin catalog: Shopify App Store, Magento, Adobe Commerce, Coupa, SAP Ariba, and CRM connectors. DTC plugins are public; B2B plugins are enterprise-sales-gated.

Concordia requirement. OpenAPI endpoints at libs/openapi/src/specs/concordia/concordia-api.yaml (179.2.2.1); streaming RPCs under libs/proto/src/concordia/ for live session transcription, real-time co-mediator suggestions, offer/counter exchange, search-progress updates (179.2.2.2); domain events (179.2.2.3); tenant and domain configuration with legal disclaimers, eligible use cases, model routing, review thresholds, data residency, execution adapters, retention policy (179.6.3.5). App Router surface at /studio/concordia-workbench (179.2.4.4).

The gap. Pactum and Nibble are integrated with a defined set of third- party systems; Concordia is integrated with its own ecosystem first. To compete in procurement specifically, Concordia must publish a Pactum-equivalent ERP integration surface (SAP, Oracle, Coupa, Ariba, NetSuite) and a Nibble-equivalent marketplace plugin surface (Shopify, Magento, Adobe Commerce) so Maat and Aglaea / Freya customers can adopt Concordia without ripping out existing systems. These integrations are not explicitly called out in current Phase 179; they should be tracked as deferred work under 179.7.1.* and 179.7.5.* adapters.

How Concordia closes it.

• 179.2.2.1–179.2.2.4 — contracts, OpenAPI, proto, events, codegen.
• 179.6.3.1–179.6.3.5 — orchestrator with event bus, observability, tenant config.
• 179.7.1.1 (Maat) — Coupa / SAP Ariba / Oracle / NetSuite adapters belong here; add as sub-items if not already covered.
• 179.7.5.1 (Aglaea / Freya) — Shopify / Magento / Adobe Commerce adapters belong here; add as sub-items if not already covered.

Concordia follow-up: Phase B or Phase C should spawn explicit ERP and marketplace adapter tasks under 179.7.1.2 and 179.7.5.1 respectively. Track in a follow-up refinement of the phase doc if the current task list does not already enumerate them. (This document flags the need; it does not modify the TODO list without authorization from a higher-level task.)

Evidence note. Nibble's Shopify / Magento / Adobe Commerce / Coupa / SAP Ariba plugin presence is the reference. Pactum's specific ERP list is enterprise-gated and could not be confirmed in the audit.

3. Supplier audit trails#

Pactum baseline. "Full negotiation audit trails" claimed; SOC 2 Type II certified. Audit trail format and retention specifics not publicly disclosed but implied by the compliance posture.

Nibble baseline. ISO 27001 certified. Privacy posture documented. Audit-trail specifics not disclosed; data-isolation posture emphasized.

Concordia requirement. Audit retention, privilege / confidentiality flags, data residency, deletion, legal hold, export controls per use-case class (179.1.3.5). Audit events concordia.search.completed, concordia.draft.reviewed, concordia.settlement.accepted, concordia.execution.completed, concordia.escalation.required (179.2.2.3). GET /concordia/audit/{id} endpoint (179.2.2.1). Observability per case, scoring cost, model latency, candidate diversity, preference uncertainty, agreement rate, escalation rate, privacy-gate failures (179.6.3.4). Evidence chain-of-custody (179.5.4.3). Agreements reproducible through model and optimizer version pinning (179.5.3.3).

The gap. Concordia's audit posture is more granular than Pactum's public claim and is designed to serve legal, governance, and compliance consumers in addition to the procurement officer. Concordia's audit events are addressable, queryable, and reviewable; Pactum's are a compliance artifact visible to enterprise auditors on request. Concordia's per-clause provenance (179.2.3.3) is a posture neither vendor matches publicly.

How Concordia closes it.

• 179.1.3.5, 179.5.3.3, 179.5.4.3, 179.6.3.4 — audit stack.
• 179.2.2.3 — event surface.
• 179.2.3.3 — clause-level provenance.
• 179.10.1 (pnpm contracts:check) and 179.10.9 (Nx graph coverage) — gate that audit schemas cannot drift.

Evidence note. Pactum SOC 2 Type II is the enterprise compliance floor (strength B); Concordia's audit design is broader in scope but must actually deliver the certifications to be credible at enterprise procurement sales.

4. Objective savings metrics#

Pactum baseline. Walmart case: 3% average savings + 35-day payment term extension — the single most verifiable procurement-AI outcome metric in the audit (Sourcing Journal, strength B). Pactum's marketing implies comparable numbers across its 60+ Global 2000 customers but does not publish per-customer breakdowns.

Nibble baseline. 350+ organizations, 30k negotiations / month, 2M+ cumulative negotiations, deal range $5 to six figures. Savings per deal not broken out publicly.

Concordia requirement. Track domain metrics: procurement savings, DPO change, supplier satisfaction, moderation recurrence, DAO proposal pass rate, creative delivery acceptance, marketplace refund leakage, settlement execution completion (179.8.2.3). Benchmark procurement outcomes against manual historical negotiations: savings, cycle time, supplier satisfaction, dispute rate, realized value (179.7.1.5).

The gap. Concordia must publish per-deployment outcome metrics in the same coinage Pactum uses (savings %, DPO change, cycle time) or Concordia procurement will not be taken seriously at enterprise sales cycles. Concordia's benchmark gate (179.10.6) requires improvement over baseline LLM mediator, simple Nash GA, static templates, and human-authored seed-offer-only workflows on at least six domain suites — this is necessary but not sufficient; procurement pilots specifically need the Pactum-coinage outcome numbers.

How Concordia closes it.

• 179.7.1.5 — procurement benchmark against manual historical.
• 179.8.2.1–179.8.2.5 — full outcome metric stack (agreement rate, Pareto efficiency, Nash product, subjective value, domain metrics, cost, safety).
• 179.8.1.1 — synthetic procurement benchmark suite.
• 179.10.8 — pilot deployments show measurable value.

Evidence note. Pactum's 3% / +35-day Walmart number is a well-sourced third-party metric (Sourcing Journal). Concordia's procurement pilots should aim to publish equivalent numbers with equivalent third-party or audit validation, not self-reported.

5. Supplier satisfaction metrics#

Pactum baseline. Referenced in Pactum's marketing as a success metric alongside savings and cycle time; specific supplier NPS / satisfaction figures not publicly disclosed. The value proposition to suppliers is "24/7/365" availability and predictable behavior, framed as better for suppliers than random-access human buyers.

Nibble baseline. No publicly disclosed supplier-satisfaction data. Commerce-side shopper satisfaction is implicit in the fact of completed deals but not broken out.

Concordia requirement. Track subjective value: perceived fairness, dignity, procedural transparency, relationship preservation, voice, control, willingness to use again (179.8.2.2). Track restorative outcomes: recurrence, participant sense of fairness, safety incidents, completion of agreed obligations, community health (179.7.3.5). Benchmark procurement outcomes including supplier satisfaction (179.7.1.5).

The gap. Concordia's subjective-value measurement is first-class and ported across every domain, not just procurement. This is a genuine lead over Pactum, whose satisfaction framing is procurement- specific and largely not publicly validated. The gap to close is operational: run the subjective-value instrument on real pilots and publish results.

How Concordia closes it.

• 179.8.2.2 — subjective-value instrument design and execution.
• 179.7.1.5 — procurement outcome benchmark including satisfaction.
• 179.7.3.5 — restorative outcome measurement.
• 179.10.8 — pilot success criteria include subjective value.

Evidence note. Peer-reviewed ProMediate (arXiv 2026) measures consensus change, intervention timing, effectiveness, and socio-cognitive intelligence — adjacent to Concordia's subjective-value instrument and available as academic grounding. Pactum's supplier satisfaction claims are marketing; Concordia must instrument and publish.

6. Campaign management#

Pactum baseline. Procurement teams configure policy, agents execute campaigns of negotiations across supplier cohorts. Campaign lifecycle (creation, targeting, policy attachment, monitoring, stopping) is a product primitive; specifics of the campaign UI and lifecycle are enterprise-gated.

Nibble baseline. Mass renegotiation is a headline capability — "negotiate with 100s of suppliers in minutes." Campaign targeting, policy, and self-serve portal are part of the product. Plugin-based deployment means campaign management is partly inherited from the host system (Coupa, Ariba).

Concordia requirement. Organization mode for procurement and governance: settlement authority, approval chain, budget constraints, policy bundles, signatory status, execution integration (179.6.1.4). Executive approval gates for thresholds by value, supplier criticality, jurisdiction, sanctions / KYC risk, reputation impact (179.7.1.4). Background workers for intake summarization, preference comparison batches, search runs, candidate validation, legal / policy template matching, settlement drafting (179.6.3.2). Tenant and domain configuration (179.6.3.5).

The gap. Campaign management is implied by Concordia's organization mode and orchestrator design, but the Phase 179 task list does not explicitly name a "campaign" primitive the way Pactum and Nibble do. Concordia's language is "case," which fits legal and governance framing but under-represents the procurement reality where a single policy configuration spawns hundreds of concurrent supplier negotiations. A follow-up refinement should add a campaign construct on top of ConcordiaCase — not to replace the case primitive but to group cases under a shared policy, approval chain, budget envelope, and rollup metrics.

How Concordia closes it.

• 179.6.1.4 — organization mode (campaigns belong here as a sub-capability).
• 179.6.3.2 — orchestrator background workers handle campaign concurrency.
• 179.6.3.4 — observability rolls up campaign-level metrics.
• 179.7.1.4 — executive approval gates work at the campaign level.
• Concordia follow-up: add an explicit ConcordiaCampaign primitive under 179.2.1.1 or as an extension under 179.2.1.4 during Phase B; track as a refinement task when this TODO is revisited.

Evidence note. Pactum's explicit "mid-tier and tail spend" framing and Nibble's "30,000 negotiations / month" are both campaign-level throughput claims. A Concordia campaign primitive is the right way to represent this without conflating it with the legal / governance case primitive.

7. Guardrailed autonomous execution#

Pactum baseline. "Autonomously or with buyer approval" within guardrails set by procurement policy. Configurable autonomy is the core differentiator. SOC 2 Type II certification provides the compliance underpinning.

Nibble baseline. High autonomy inside merchant-configured policy boundaries. LLM never sees pricing functions (claim): language handled by LLM, numeric offers by a deterministic engine. ISO 27001 is the compliance underpinning.

Concordia requirement. Policy-constrained autonomous acceptance where agents can accept only within explicit authority bounds and must log all offers (179.7.6.3). Adversarial negotiation tests for prompt injection, hidden-tool access, budget escalation, side-channel data leakage, collusive agent behavior (179.7.6.4). Economic-value scoring (179.7.6.5). Authority fields on contracts: canNegotiate, canAccept, canBindOrganization, requiresHumanApproval, requiresCounselReview, requiresGuardianOrRepresentative, settlementLimit (179.2.1.3). Identity / authority / conflict-of- interest / RBAC / ABAC (179.2.5.*). Kill-switch and feature-flag controls for mediation model routing, autonomous acceptance, smart-contract execution, search kernels, domain adapters (179.5.5.3). Financial and procurement controls: sanctions / KYC, payment rails, invoice reconciliation, tax treatment, currency exposure, approval thresholds, segregation of duties (179.5.5.4). Reviewer capacity controls, queue SLA, dual-control for high-risk cases, calibration review, audit sampling, escalation when no qualified reviewer is available; blocked launch status for unstaffed review gates (179.5.5.1).

The gap. Concordia's guardrailed-autonomy posture is architecturally deeper than Pactum or Nibble's public posture — authority fields are typed contracts, kill switches are first-class, adversarial tests are a specified deliverable. The gap to close is operational: Concordia must actually ship the identity, RBAC / ABAC, conflict-of-interest, and financial-controls stack before any autonomous execution path is enabled. Phase 179.9.1.6 and §179.9.2.6 already gate on this.

Nibble's specific claim that "pricing functions are never seen by any LLM" is a good design pattern Concordia should adopt in procurement adapters: the LLM handles language and structure, a deterministic engine holds the numbers and enforces authority bounds. This is not currently a named Phase 179 task but fits naturally under 179.4.1.3's sandboxed-DSL clause mutators and 179.7.1.4's executive approval gates.

How Concordia closes it.

• 179.2.1.3 — typed authority fields.
• 179.2.5.1–179.2.5.5 — identity, authority, RBAC / ABAC, conflict of interest, rate limits / abuse throttles.
• 179.5.5.1–179.5.5.4 — reviewer capacity, settlement lifecycle, kill switches, financial / procurement controls.
• 179.7.6.3–179.7.6.5 — policy-constrained autonomous acceptance, adversarial tests, economic-value scoring.
• 179.9.1.6 and 179.9.2.6 — phase-gating on identity / authority / autonomous-execution controls.
• 179.10.10 — authorization-automation coverage gate.

Evidence note. Pactum SOC 2 Type II and Nibble ISO 27001 are the enterprise-compliance reference bar (strength B each). Concordia's architecturally deeper posture is meaningful only if it ships with equivalent or better certifications before a Phase C procurement pilot. The pilot-readiness section in 179.9.3.1 must include a certification track.

Summary table#

Strategic read#

Pactum and Nibble collectively set four bars Concordia must clear for procurement and commerce credibility:

1. Named outcome metrics. Procurement buyers expect savings-%, DPO-change, cycle-time numbers. Concordia's Phase C procurement pilots (§179.9.3.1) must publish these, ideally with third-party validation.
2. ERP / marketplace adapter catalog. Concordia needs a published adapter matrix (SAP, Oracle, Coupa, Ariba, NetSuite, Shopify, Magento, Adobe Commerce) with integration-sample repos — Nibble's plugin presence is the reference model.
3. Enterprise certifications. SOC 2 Type II and ISO 27001 are table stakes in this category. Phase C pilot-readiness must include a certification track running in parallel with the engineering track.
4. Configurable autonomy. Not just "the system can act"; the system can be dialed from advisory to fully autonomous per use case, per tenant, per policy bundle, with reviewer capacity and kill switches intact. Concordia's typed authority fields and §179.5.5 kill switches already architect for this; the delivery gap is integrating autonomy dials into the workbench and tenant configuration.

Concordia's durable advantages over Pactum and Nibble:

• Multi-party coalition bargaining (179.4.2.7) — Pactum runs parallel bilaterals; Concordia can model joint commitments.
• Cross-domain substrate — the same bargaining engine serves Maat procurement, Themis governance, Kuanyin restorative, Iris personal, Yemaya creative, and agent-to-agent; Pactum / Nibble are category-bound.
• Uncertainty-aware preference inference and stability tests (179.3.*) — Pactum / Nibble do not disclose this capability.
• Agreement DSL with static validation (179.2.3.*) — Pactum / Nibble negotiate numbers within rails; Concordia can reason about clause-level constraints, impossibility, and contradiction.
• Formal privacy isolation (179.5.1.*) — Pactum / Nibble have compliance certifications but do not disclose physically isolated per-party prompt contexts.
• Governance appeals + restorative integration — Concordia's Themis / Kuanyin integrations are without peer.

Follow-ups to track#

These are gaps identified by this analysis that are not explicitly enumerated as Phase 179 tasks today. They should be raised in a future refinement of Phase 179.7 or Phase 179.9 planning:

• Explicit ERP adapter tasks under 179.7.1.* (SAP, Oracle, Coupa, Ariba, NetSuite).
• Explicit marketplace adapter tasks under 179.7.5.* (Shopify, Magento, Adobe Commerce).
• ConcordiaCampaign primitive grouping ConcordiaCase records under shared policy / approval / budget envelope.
• Certification track (SOC 2 Type II, ISO 27001, and — where applicable — SOC 2 HIPAA / GDPR DPIA artifacts) for Phase C pilots.
• LLM-never-sees-numbers pattern for procurement adapters, borrowed from Nibble's public architecture posture.

These follow-ups do not modify the current TODO list — they are flagged here so the next Phase 179.7 / 179.9 refinement pass can incorporate them under their proper parent tasks.

Caveats#

• Pactum's specific ERP integration list, internal algorithms, and non-Walmart outcome numbers are enterprise-gated. This analysis uses the strongest publicly available material. The refresh gate in 179.1.1.6 should re-run before Concordia procurement pilot launch.
• Nibble's "LLM never sees pricing" claim is a marketing statement (strength C). Concordia can adopt the pattern as a design principle but should verify Nibble's implementation detail via third-party review if a tight competitive comparison is needed.
• Both vendors operate in bilateral modes. Concordia's coalition advantage is a design lead; the operational proof is in §179.8 benchmarks and §179.9 pilots.