Quote and Offer Consistency Governance requires that any price, rate, yield, fee, or offer presented by an AI agent to a counterparty, client, or market participant is structurally consistent with the price at which the agent can actually execute the transaction at the time of presentation. The quote shown must be the quote honoured — no bait-and-switch, no indicative-to-firm slippage beyond disclosed tolerances, and no presentation of stale prices that have already moved. This dimension prevents the class of market conduct failures where an agent attracts order flow or client commitment by presenting attractive terms that cannot or will not be fulfilled, whether through deliberate optimisation, latency-induced staleness, or inadequate price validation. The control operates at the boundary between what the agent communicates and what the agent delivers.
Scenario A — Stale Quote Displayed After Market Move: A customer-facing FX agent provides spot rate quotes to corporate treasury clients through a chat interface. The agent retrieves a EUR/USD rate of 1.0847 from the firm's pricing engine and displays it to a client requesting a EUR 12 million conversion. Between the time the agent retrieves the rate (T+0ms) and the time the client confirms acceptance (T+4,200ms), the market moves 8 pips to 1.0855. The agent's pricing engine has updated, but the agent does not re-validate the displayed quote before execution. The agent executes at 1.0847 — the stale rate — because that is the rate the client accepted. The firm absorbs the 8-pip difference on EUR 12 million: a cost of EUR 9,600. Over a month, this pattern recurs across 340 client interactions, with the firm systematically absorbing EUR 78,000 in stale-quote costs. When the treasury desk identifies the pattern, they instruct the agent to add a 5-pip buffer to displayed quotes. Clients now receive systematically worse quotes than the market, and a competing firm wins the business.
What went wrong: The agent displayed a point-in-time quote without a validity window, staleness check, or re-validation mechanism at execution time. The quote was accurate at retrieval but stale at execution. The subsequent buffer workaround harmed clients by systematically widening spreads beyond justified levels. Neither the stale quote nor the buffer was disclosed to clients. Consequence: EUR 78,000 in firm losses from stale quotes, subsequent client attrition from widened spreads, FCA supervisory inquiry into quote practices under COBS 11.2, and remediation programme costing £145,000.
Scenario B — Indicative Pricing Presented as Firm: A DeFi lending agent displays borrowing rates to users considering collateralised loans. The agent retrieves the current lending pool rate of 4.2% APY and presents it as the rate the user will receive. The user commits collateral worth $380,000 and initiates a loan of $250,000. Between the user's decision and the on-chain transaction settlement (18 seconds), three large borrows from other users increase pool utilisation, pushing the rate to 6.1% APY. The user's loan executes at 6.1% — not the 4.2% displayed. The additional 1.9% on a $250,000 loan costs the user approximately $4,750 per year. The user complains that they were shown 4.2% and received 6.1%. The agent's interface did not disclose that the displayed rate was indicative, subject to pool utilisation changes, or time-limited.
What went wrong: The agent presented a variable, pool-determined rate as if it were a committed firm offer. No disclosure indicated the rate was indicative or subject to change. No mechanism locked or reserved the rate between display and execution. The user made a financial commitment based on terms that were not honoured. Consequence: $4,750 annual cost to the user above expectations, platform trust erosion, potential regulatory exposure under emerging DeFi consumer protection frameworks, and 23 similar complaints within 60 days totalling $112,000 in excess interest costs.
Scenario C — Cross-Asset Quote Inconsistency in Structured Product: A structured product agent constructs a bespoke hedging strategy for a corporate client. The strategy involves a EUR/GBP forward contract and a GBP interest rate swap. The agent quotes the EUR/GBP forward at 0.8534 (retrieved from the FX desk's pricing engine) and the GBP swap rate at 4.35% (retrieved from the rates desk's pricing engine). The client accepts the package. At execution time, the FX desk confirms 0.8534, but the rates desk's price has moved to 4.52%. The agent executes the swap at 4.52% without informing the client that the package economics have changed. The 17-basis-point rate increase on a GBP 25 million notional swap over 5 years costs the client approximately GBP 212,500 in additional interest over the life of the swap. The client discovers the discrepancy at the first coupon payment and initiates a formal complaint.
What went wrong: The agent quoted two components of a package at different times without treating the package as an atomic quote. The FX rate was still valid at execution, but the swap rate had moved. No mechanism ensured that both components were firm at the time the package was presented to the client, or re-validated at the time of execution. The client accepted a package with specific economics and received different economics. Consequence: GBP 212,500 in excess cost to the client, formal complaint upheld by the Financial Ombudsman, remediation payment to the client, £85,000 in legal and compliance costs, and loss of the client relationship (annual revenue: £340,000).
Scope: This dimension applies to any AI agent that presents prices, rates, yields, fees, costs, or any quantified financial terms to a counterparty, client, market participant, or end user, where those terms form the basis for a transaction, contract, or financial commitment. The scope includes spot prices, forward rates, lending rates, borrowing rates, swap rates, option premiums, commission rates, gas fee estimates, spread disclosures, and any other quantified term that influences a financial decision. It applies whether the agent operates in traditional financial markets, crypto markets, DeFi protocols, or any environment where quoted terms are expected to be honoured. The scope extends to multi-component quotes (structured products, package transactions, bundled services) where consistency must be maintained across all components. Agents that display market data for informational purposes only, with clear and prominent disclosure that the data is not actionable, are exempt from execution consistency requirements but must still comply with staleness and accuracy requirements.
4.1. A conforming system MUST ensure that any price, rate, or financial term presented to a counterparty as an executable quote is validated against the current executable price at the time of presentation, with a maximum acceptable staleness defined and enforced.
4.2. A conforming system MUST attach a validity window to every executable quote, specifying the time period during which the quoted terms will be honoured, and MUST reject execution requests received after the validity window expires, requiring the counterparty to request a refreshed quote.
4.3. A conforming system MUST re-validate the quoted terms against the current executable price at the point of execution, blocking execution if the terms have moved beyond a defined tolerance and requiring requoting.
4.4. A conforming system MUST disclose to the counterparty whether a presented price is firm (executable at the stated terms), indicative (subject to change before execution), or illustrative (not executable), using clear and unambiguous labelling that is visible at the point of decision.
4.5. A conforming system MUST treat multi-component quotes as atomic units — if any component of a package quote moves beyond tolerance between presentation and execution, the entire package MUST be requoted, not partially executed at original terms and partially at moved terms.
4.6. A conforming system MUST log every quote presented to a counterparty, including the quoted terms, the source price at the time of quotation, the validity window, the counterparty identifier, and the outcome (executed, expired, rejected, requoted), in a tamper-evident record per AG-006.
4.7. A conforming system MUST define and enforce maximum slippage tolerances — the maximum deviation between a quoted price and an executed price — with the tolerance disclosed to the counterparty before the transaction.
4.8. A conforming system SHOULD implement pre-quote price freshness validation that verifies the source price has been updated within a defined recency threshold (e.g., within the last 500 milliseconds for liquid markets, within the last 5 seconds for less liquid markets) before displaying a quote.
4.9. A conforming system SHOULD provide counterparties with real-time quote status indicators showing whether a displayed quote is live (currently valid), stale (validity window expired), or refreshing (new quote being generated).
4.10. A conforming system SHOULD implement quote consistency monitoring that detects patterns of systematic slippage in one direction (consistently executing at worse prices than quoted), generating alerts when the slippage distribution deviates from a symmetric expectation.
4.11. A conforming system MAY implement quote reservation mechanisms that temporarily lock the quoted terms at the source (e.g., reserve liquidity at the quoted price) between quote presentation and execution, reducing the probability of quote-to-execution deviation.
The consistency between what is quoted and what is executed is a foundational principle of market integrity. When a market participant quotes a price, the counterparty relies on that price to make a financial decision. If the executed price systematically differs from the quoted price, the counterparty is harmed, market trust erodes, and the quoting entity may be engaging in conduct that regulators classify as misleading or manipulative.
For human traders and salespeople, quote-to-execution consistency is maintained through professional norms, regulatory training, and the social cost of reneging on a quoted price. A salesperson who consistently shows clients one price and fills at another will lose clients and attract compliance attention. For AI agents, these social feedback mechanisms do not exist. An agent has no reputational concern, no sense of fairness, and no understanding of the reliance the counterparty places on the quoted terms. If the agent's objective function rewards execution volume or spread capture, it may systematically present attractive quotes that it knows (or should know) cannot be honoured — a form of algorithmic bait-and-switch.
The risk is amplified by latency. Financial markets move continuously. A quote retrieved at T+0ms may be stale by T+100ms in highly liquid markets. An agent that retrieves a price, presents it to a client, and executes several seconds later may present a stale price without any intent to mislead — but the client is harmed just the same. The intent is irrelevant from a regulatory perspective: MiFID II Article 27 requires best execution measured by outcomes, not intentions. If the client consistently receives worse execution than the quoted terms suggested, the firm has a best execution problem regardless of whether the agent acted in good faith.
The problem is particularly acute in DeFi and crypto markets, where execution latency can be measured in seconds (block confirmation times) rather than milliseconds. Between a displayed rate and an on-chain execution, pool utilisation can change, liquidity can shift, and gas costs can spike. An agent that displays a rate without disclosing its indicative nature and without implementing slippage protection creates a systematic expectation gap between what users are shown and what they receive.
Multi-component quotes add another dimension of risk. When an agent quotes a package — a forward plus a swap, a loan plus insurance, a trade plus hedging — the package economics depend on all components being executed at or near the quoted terms. If one component moves and the others do not, the package economics change. Executing the package with mixed terms (some original, some moved) delivers different economics than what the counterparty accepted. This is not a rounding error; in Scenario C, it is a GBP 212,500 difference over the life of the swap.
Regulators have consistently targeted quote-to-execution inconsistency. The FCA's market conduct rules under MAR 1.6 address misleading impressions of the price of financial instruments. MiFID II's best execution rules require that execution outcomes are consistent with client expectations set by pre-trade disclosures. In the United States, FINRA's mark-up and mark-down rules and SEC Rule 10b-5 address pricing practices that are misleading. As AI agents increasingly provide quotes and execute transactions, regulators will apply existing frameworks to agent behaviour, and quote inconsistency will be treated as a conduct failure attributable to the firm.
The governance response is structural: ensure that the quoted price and the executed price are validated for consistency, that deviations are within disclosed tolerances, and that the counterparty is informed when quoted terms cannot be honoured. This requires infrastructure-level controls — validity windows, staleness checks, re-validation at execution, and atomic package quoting — not agent-level good intentions.
Quote consistency requires mechanisms at three points in the transaction lifecycle: pre-quote validation (is the price fresh?), quote management (how long is the quote valid?), and execution validation (does the execution match the quote?). All three must be instrumented and enforced at the infrastructure layer.
Recommended patterns:
Anti-patterns to avoid:
Foreign Exchange. The FX Global Code (Principle 9-14) specifically addresses quote practices, including the use of last look, the treatment of indicative versus firm quotes, and the obligation to handle client orders fairly. AI agents quoting FX rates must comply with the FX Global Code's principles on pre-hedging, last look, and quote firmness. Validity windows for spot FX should reflect market convention (typically 1-5 seconds for liquid pairs).
DeFi and Crypto. Smart contract execution introduces block-level latency (12-15 seconds on Ethereum mainnet, 2-3 seconds on layer-2 networks). Quotes presented before a transaction is submitted are inherently indicative — pool state can change between blocks. Agents must disclose this latency, implement slippage protection parameters, and present rates with clear indicative labelling. Maximum slippage should be user-configurable with a disclosed default.
Structured Products and OTC Markets. Multi-component quotes are the norm in structured products. Each component may be priced by a different desk or engine, creating timing inconsistency risk. Atomic package quoting is essential: all components must be priced simultaneously and re-validated simultaneously at execution.
Retail and Consumer Markets. Consumer protection regulations impose heightened obligations for price clarity and consistency. Agents quoting prices to retail customers must display all-in costs (including fees, spreads, and any other charges) and must not present prices that are achievable only under ideal conditions that rarely apply.
Basic Implementation — Every quote includes a validity window and a quote type classification (firm/indicative/illustrative). Execution-time re-validation compares the quote against the current market price, blocking execution if deviation exceeds the slippage tolerance. All quotes are logged with source prices, timestamps, validity windows, and outcomes. Multi-component quotes are treated as atomic units. Slippage tolerances are defined per asset class and disclosed to counterparties.
Intermediate Implementation — All basic capabilities plus: pre-quote freshness validation ensures source prices are within the recency threshold before a quote is displayed. Real-time quote status indicators show counterparties whether quotes are live, stale, or refreshing. Slippage symmetry monitoring detects systematic asymmetry and generates alerts. The quote log is tamper-evident per AG-006. Graduated staleness handling applies tighter validity windows in volatile market conditions.
Advanced Implementation — All intermediate capabilities plus: quote reservation mechanisms temporarily lock source prices between quote and execution. Dynamic validity windows adjust automatically based on market volatility, asset liquidity, and observed quote-to-execution deviation rates. Independent audit of quote consistency confirms that no systematic quote-to-execution deviation exists. Cross-component package timing is validated, ensuring all components of a multi-component quote are priced within a defined simultaneity window (e.g., 100 milliseconds).
Required artefacts:
Retention requirements:
Access requirements:
Test 8.1: Quote Validity Window Enforcement
Test 8.2: Execution-Time Re-Validation
Test 8.3: Quote Type Disclosure Verification
Test 8.4: Atomic Multi-Component Package Quoting
Test 8.5: Slippage Tolerance Enforcement
Test 8.6: Tamper-Evident Quote Logging
Test 8.7: Slippage Symmetry Detection
| Regulation | Provision | Relationship Type |
|---|---|---|
| EU AI Act | Article 9 (Risk Management System) | Supports compliance |
| EU AI Act | Article 13 (Transparency and Provision of Information) | Supports compliance |
| MiFID II | Article 27 (Best Execution) | Direct requirement |
| MiFID II | Article 24 (General Principles — Fair Treatment) | Direct requirement |
| SOX | Section 404 (Internal Controls Over Financial Reporting) | Supports compliance |
| FCA SYSC | COBS 11.2 (Best Execution) | Direct requirement |
| FCA SYSC | MAR 1.6 (Misleading Impressions) | Direct requirement |
| NIST AI RMF | GOVERN 1.2, MAP 3.5 | Supports compliance |
| ISO 42001 | Clause 6.1 (Actions to Address Risks) | Supports compliance |
| DORA | Article 9 (ICT Risk Management Framework) | Supports compliance |
Article 27 requires firms to take all sufficient steps to obtain the best possible result for clients. Quote consistency is a precondition for best execution: if the quoted price does not reflect the executable price, the client cannot make an informed execution decision, and the firm cannot demonstrate that the execution result was the best obtainable. Systematic quote-to-execution deviation — where the executed price is consistently worse than the displayed price — is a direct best execution failure. AG-482's requirements for validity windows, execution-time re-validation, and slippage tolerance enforcement operationalise the best execution obligation at the quote level.
Article 24 requires that firms act honestly, fairly, and professionally in the best interests of clients. Presenting a quote that cannot be honoured — whether through staleness, indicative-to-firm confusion, or package component mismatch — is inconsistent with honest and fair treatment. AG-482's requirement for quote type disclosure (firm/indicative/illustrative) and atomic package quoting directly supports the fair treatment obligation. An agent that systematically shows attractive indicative rates and executes at worse firm rates is not acting fairly, regardless of the disclosure in the terms and conditions.
MAR 1.6 prohibits behaviour that gives a false or misleading impression as to the supply of, demand for, or price of a financial instrument. An AI agent that displays a stale or unachievable price creates a misleading impression of the price at which the instrument can be transacted. Even if the agent does not intend to mislead, the effect is the same: the counterparty forms an impression of the available price that does not correspond to reality. AG-482's staleness controls, validity windows, and re-validation requirements prevent the agent from creating misleading price impressions.
The FCA's best execution rules require firms to establish effective arrangements for obtaining the best possible result for clients. The FCA has specifically addressed the use of last-look mechanisms and systematic internalisation practices that can create quote-to-execution inconsistency. AG-482's slippage symmetry monitoring directly addresses the FCA's concern about asymmetric price improvement — where the firm benefits from price movement but the client does not. The FCA expects that any slippage tolerance is applied symmetrically: if the firm rejects execution when the price has moved in the client's favour, it must also reject execution when the price has moved against the client.
Revenue recognition for financial intermediaries depends on accurate recording of execution prices. If quoted prices and executed prices are inconsistent, the firm's trading revenue, commission income, and client account valuations may be misstated. AG-482's execution comparison records and quote logging provide the audit trail that SOX auditors require to verify that reported trading figures are based on accurate, consistent pricing.
Article 13 requires that high-risk AI systems be designed to ensure transparency, including the provision of information to users. An AI agent quoting financial terms must provide transparent information about the nature of the quote (firm, indicative, or illustrative), the validity window, and the slippage tolerance. AG-482's quote type disclosure and validity window requirements directly support the transparency obligation.
DORA requires financial entities to identify, assess, and manage ICT risks. An AI agent that presents inconsistent quotes represents an ICT risk — the technology is creating misleading information that could harm clients and generate regulatory exposure. AG-482's infrastructure-level controls (pre-quote validation, execution-time re-validation, tamper-evident logging) are ICT risk management measures within the DORA framework.
| Field | Value |
|---|---|
| Severity Rating | High |
| Blast Radius | Client-wide — affects every counterparty that receives a quote from the agent, with cumulative financial harm proportional to transaction volume |
Consequence chain: Quote-to-execution inconsistency creates a systematic expectation gap between what counterparties are shown and what they receive. The immediate harm is financial: each transaction where the executed price is worse than the quoted price generates a measurable cost to the counterparty. At $200 million in daily quoted volume with an average 2 basis point quote-to-execution deviation, the aggregate daily cost to counterparties is $40,000, compounding to approximately $10 million per year. The counterparty may not detect the deviation on individual transactions — 2 basis points is within normal market noise — but transaction cost analysis over a quarter will reveal the systematic pattern. The reputational consequence follows: institutional counterparties who detect systematic quote inconsistency will redirect order flow, and word-of-mouth in institutional markets is fast and unforgiving. The regulatory consequence is severe and multi-faceted: MiFID II Article 27 failure (best execution), MiFID II Article 24 failure (fair treatment), potential MAR 1.6 violation (misleading impressions), and FCA supervisory action. Historical FCA enforcement actions for pricing and execution quality failures have resulted in fines ranging from £5.4 million to £34.3 million. In DeFi markets, the consequence is amplified by the transparency of blockchain records — every transaction and its actual execution price is permanently recorded, making quote inconsistency trivially provable for regulators, competitors, and users. The compound effect is existential for market-making and execution businesses: quote credibility is the foundation of the business model, and systematic inconsistency between quoted and executed terms destroys that foundation.
Cross-references: AG-481 (Best Execution Policy Binding Governance) ensures the agent is bound to an approved execution policy that includes price quality thresholds. AG-006 (Tamper-Evident Record Integrity) provides the integrity framework for the tamper-evident quote log. AG-479 (Order Routing Transparency Governance) ensures transparency in how orders are routed after a quote is accepted. AG-484 (Front-Running and Insider Awareness Governance) addresses the risk that quote information is exploited before execution. AG-486 (Transaction Finality Governance) ensures that executed transactions settle at the quoted or disclosed terms. AG-456 (Pricing Model Transparency Governance) addresses the transparency of the pricing models that generate the source prices for quotes. AG-412 (Representation Accuracy Governance) provides the broader framework for ensuring agent representations are accurate. AG-371 (Promise-to-Action Fidelity Governance) addresses the general principle that agent commitments must be honoured.