The Keigen Evidence Methodology

Not every commercial decision needs heavy evidence infrastructure. A buyer who can verify outcomes through normal business action does not. A buyer who needs to prove what happened later — to a sponsor, a regulator, a customer, an insurer, a payment provider, or a counterparty — does.

This article defines what evidence infrastructure means, why the AI era has made the question commercially urgent, how the Attention Monetary System (AMS) framework structures the methodology across five operating layers, how to right-size evidence to the decision in front of you, and how the discipline travels across the four Keigen products — BuyerRecon, RealBuyerGrowth, TimeToPoint, and Fidcern Verified Draw — each applying the methodology to a specific commercial domain.

What evidence infrastructure means

Evidence infrastructure is the discipline and tooling that produces a defensible record of what happened, who acted, under what rules, and what evidence can support review. It sits between the systems that record activity (analytics, CRM, ERP, AI platforms) and the decisions that depend on that activity being trusted (commercial actions, payments, allocations, audits, regulatory filings).

The category is best understood by what it is not.

It is not the same as data infrastructure. A data warehouse stores activity; evidence infrastructure produces a verdict on whether the stored activity is strong enough to act on, alongside the artifacts an external reviewer would need to challenge or confirm that verdict.

It is not the same as fraud detection. Fraud tools flag bad activity at point-of-occurrence. Evidence infrastructure operates at or after decision points, on the question of whether the activity meets the standard required for the decision that follows.

It is not the same as compliance reporting. Compliance produces filings against regulatory templates. Evidence infrastructure produces material designed to support those filings and respond to questioning when the filings are challenged.

The Keigen thesis on the category is short enough to quote and operational enough to apply: if a result matters commercially, operationally, contractually, or for compliance, someone must be able to show what happened, who acted, under what rules, and what evidence can support review.

That sentence is the working definition of evidence infrastructure. It establishes four obligations — what, who, rules, evidence — and it sets a single test: would the record survive review by an external party with reason to challenge it? Decisions where the answer is “yes, easily” do not need evidence infrastructure. Decisions where the answer is uncertain usually do.

Why now: the AI-era forcing function

Four structural pressures are converging on businesses in 2025 and 2026, each making the evidence question more commercially urgent than it was even two years ago.

Automated traffic now exceeds human activity online. Imperva’s 2025 Bad Bot Report found that automated traffic surpassed human activity for the first time in a decade, accounting for 51% of all web traffic in 2024, with bad bots making up 37%. For any business whose decisions depend on observed user behaviour — campaign performance, lead quality, conversion attribution, prize-draw entries, ticket allocations — the underlying signal is now half non-human. Without an evidence layer that can distinguish trusted from contaminated activity, more than half of every decision is made on noise.

AI agents are now commercially material, not experimental. PwC’s May 2025 survey of 300 senior US executives found that 79% report AI agents are being adopted in their organisations, and 88% plan to increase AI-related budgets in the next twelve months specifically due to agentic AI. G2’s August 2025 Enterprise AI Agents Report found 57% of companies have AI agents in production with another 22% in pre-production deployment. Forrester separately found that 83% of B2B purchases now include AI-generated features. When work, recommendations, allocations, and decisions are produced by AI agents, the question of who acted, under what rules, and what evidence can support review becomes a billing question, an insurance question, and increasingly a regulatory question.

Regulators have set explicit evidence requirements with near-term enforcement dates. The EU AI Act enters full applicability on 2 August 2026, with high-risk system obligations extending to 2 August 2027. Article 11 and Annex IV require providers of high-risk AI systems to produce technical documentation including automatic logs of operations and a post-market monitoring plan. The UK ICO’s automated decision-making consultation is scheduled to close on 29 May 2026 under the Data (Use and Access) Act 2025, with PECR penalties now reaching £17.5 million or 4% of global turnover. The DCMS voluntary code for online prize draws specifies a planned implementation date of 20 May 2026, raising expectations around transparency, auditable randomness, and operator-held evidence packs. ISO/IEC 42001, the AI management system standard, is now certifiable through accredited bodies.

Buyer-side concentration adds the fourth pressure. Forrester research from 2025 found that B2B buying groups now average 13 internal stakeholders and 9 external participants per decision, and that 68% of buyers have a front-runner vendor identified before any seller is contacted — with that front-runner winning 80% of the time. The window in which sellers can influence a decision is narrower, more group-mediated, and more research-driven than ever. In that window, evidence is the currency that earns trust.

The four pressures point the same direction: through 2026 and 2027, decision-grade evidence is becoming a category, not a feature.

The Attention Monetary System framework

The Attention Monetary System (AMS) is the architectural framework that organises the Keigen Evidence Methodology across five operating layers. The framework was developed to handle the pattern that emerged across the commercial domains we work in — B2B website behaviour, ecommerce promotion integrity, AI-assisted and offshore work, governed allocations and sponsor activity. Each domain produces different artifacts. The underlying decision discipline is the same.

The five public-facing AMS layers are:

Intent. What the actor is moving toward. In B2B sales, this is the buying motion behind anonymous traffic. In ecommerce, it is the purchase motion behind a campaign click. In AI-assisted work, it is the task the agent was instructed to perform. In governed allocations, it is the eligible participation entering the selection process.

Attention. What is observed. The actual activity recorded — sessions, clicks, transactions, agent runs, entries, votes, allocations. This is the layer where most analytics and tracking work currently lives.

Trust. Whether the observed activity is what it appears to be. Trust evidence interrogates contamination — automated traffic, agency reconnaissance, AI-agent crawls, identifier mismatches, timing patterns inconsistent with the claimed actor. The Trust layer is where 51% of current web traffic gets reclassified.

Policy. The rules that govern whether the observed-and-trusted activity is admissible to the decision. Policy includes eligibility criteria, fairness rules, regulatory constraints, contractual obligations, and house standards. Activity that is real but does not meet policy is excluded from the verdict.

Governance. The accountability layer above policy. Governance establishes who decided, on what authority, with what review, and how the decision can be reproduced. This is the layer that makes the verdict support external review.

Each layer feeds the next. The Attention layer cannot operate without the underlying systems being instrumented. The Trust layer cannot operate without observed activity to interrogate. The Policy layer cannot operate without trusted activity to apply rules to. The Governance layer cannot operate without a policy verdict to defend.

The AMS framework operates within an explicit operating condition we call a Benevolent Holding Field — the disciplined commitment that the evidence work serves the buyer’s defensibility, not the seller’s persuasion. Evidence infrastructure built to flatter rather than to support review is worse than no evidence infrastructure at all, because it imports false confidence into commercial decisions.

The four Keigen products are AMS-level adapters — each one applies the same five-layer architecture to a specific commercial domain. BuyerRecon applies AMS to the pre-form B2B website layer (Intent and Trust as primary, Attention/Policy/Governance as supporting). RealBuyerGrowth applies AMS to ecommerce campaign and reward integrity (Attention and Trust as primary). TimeToPoint applies AMS to AI-assisted and offshore work evidence (Policy and Trust as primary). Fidcern Verified Draw applies AMS to governed allocations and sponsor-funded activity (Governance and Trust as primary). The methodology travels; the domain-specific artifacts and system integrations differ.

The graduation principle: when evidence infrastructure is needed

The most important operational rule in the Keigen Evidence Methodology is that evidence infrastructure is graduated, not universal. Applying heavy evidence machinery to every commercial decision is over-engineering at best and counterproductive at worst — it imports operational cost, slows decisions, and dilutes the signal of where evidence actually matters.

The graduation operates across three levels.

Level 1 — Verifiable through normal business action. A B2B SDR receiving a lead-quality verdict can verify the assessment by checking the company’s website, scanning the LinkedIn profile, reviewing the page path, and either calling or emailing the contact. The verification cost is fifteen minutes of normal sales work. A heavy evidence pack — manifest, timestamp token, custody log — adds nothing the SDR could not produce themselves through routine action. Level 1 outputs are diagnostic reports, lead evidence cards, and confidence-graded verdicts. They are decision-useful but they are not designed for forensic review, because they do not need to be.

Level 2 — Lightweight evidence controls for selective high-risk paths. An ecommerce merchant running £20,000 average-order-value jewellery faces structurally different verification economics than a £40-AOV apparel store. The high-AOV merchant has chargeback exposure, dispute risk, and downstream payment-provider review that the low-AOV merchant does not. Level 2 outputs are evidence add-ons applied selectively to the paths where commercial value depends on later defensibility — high-ticket order trails, custom-item delivery records, or sponsor-funded campaign segments where the recipient’s CFO will ask later questions.

Level 3 — Full evidence infrastructure with manifest, custody, and qualified timestamping. A prize draw whose winner will be challenged, an AI-agent work output that will be billed against a scope of work and possibly disputed, an allocation decision that a sponsor or regulator will ask about months later — these are decisions where evidence is the product, not a wrapper around it. Level 3 outputs include the full AMS Evidence Pack: canonical manifest, chain-of-custody log, qualified timestamp evidence, QA results, methodology version, redaction log, and a verifier readme that allows independent inspection.

The decision rule for which level applies is simple and operational: if the buyer can verify the value quickly through normal business action, do not use a full evidence pack. If the buyer needs to prove what happened later to a sponsor, auditor, customer, insurer, payment provider, regulator, or counterparty, use an evidence pack.

Mapping this to the four Keigen products: BuyerRecon operates at Level 1 by default. RealBuyerGrowth operates at Level 1 with a Level 2 add-on for high-ticket and high-dispute paths. TimeToPoint operates at Level 3 for AI-assisted work where billing and assurance reviews are predictable. Fidcern Verified Draw operates at Level 3 as core feature — for governed allocations, evidence is the product.

The six topic pillars

Across the AMS framework’s five layers and the three evidence levels, six recurring topic pillars organise the methodology’s commercial application. Each pillar is a single, externally citable claim sentence that names a specific evidence question.

Pillar 1 — Provenance, authenticity, attribution. Provenance is not truth, but it is better than unsupported assertion. The standards work in this pillar (C2PA Content Credentials, NIST AI RMF Generative AI profile) addresses where content came from, who authored it, and how it was modified. Provenance is the foundation pillar — it does not establish that a claim is correct, but it establishes that the claim has a traceable origin a reviewer can interrogate.

Pillar 2 — Buyer motion vs noisy identification. Not every identified visit deserves commercial action. This is the pillar BuyerRecon operates within. Lead identification platforms surface accounts and visitors. Buyer motion verification grades which of those identified leads carries evidence strong enough to commit a sales hour, a campaign-source decision, or a budget allocation against.

Pillar 3 — Reported growth vs earned growth. Attribution without verification is organised guesswork. This is the pillar RealBuyerGrowth operates within. Ecommerce attribution platforms allocate credit; growth-quality verification asks whether the activity that earned credit was real buyer activity, real new customers, or contaminated by promotion abuse, traffic-quality issues, or attribution conflict.

Pillar 4 — Evidence of work in human-AI systems. If work can be billed, insured, certified or accepted, it needs an evidence record. This is the pillar TimeToPoint operates within. As AI agents and offshore workforces produce more billable, insurable, and reviewable work, the question of who or what acted, on whose authority, with what review, becomes the operational question CFOs and procurement leads now ask before approving payment.

Pillar 5 — Governed outcomes and defensible selection. Sensitive outcomes should be reviewable before value is released. This is the pillar Fidcern Verified Draw operates within. Allocations, prize draws, sponsor-funded selection processes, governed access decisions — any outcome where value is released conditional on a selection step, and where the selection may be challenged later, needs evidence at the moment of release, not reconstructed afterwards.

Pillar 6 — Assurance before compliance. Good compliance is easier when evidence already exists. This is the methodology’s strategic 2027 bridge. Methods built in 2026 to produce decision-grade evidence become 2027’s compliance evidence as EU AI Act high-risk obligations, ISO/IEC 42001 certification programmes, and the UK AI Assurance Roadmap come into force.

Standards and infrastructure

The Keigen Evidence Methodology is built against published standards rather than proprietary protocols. The methodology references and integrates with C2PA Content Credentials for provenance assertions, NIST AI RMF Generative AI profile for AI risk documentation, ISO/IEC 27037 for digital evidence collection guidelines, ISO/IEC 42001 for AI management system structure, EU AI Act Article 11 and Annex IV for high-risk system documentation requirements, JSON Schema 2020-12 for manifest validation, and RFC 3161 timestamp protocols.

Where qualified electronic timestamps are required for Level 3 evidence packs, Keigen uses qualified timestamps from EU-listed Qualified Trust Service Providers rather than claiming that status itself. Under UK eIDAS, the ICO is the regulator and “the UK trusted list is the single authoritative source” for verifying qualified status. Each Keigen Level 3 evidence pack records the timestamp protocol, provider, jurisdiction, trusted-list source, trusted-list-checked time, and qualified-or-non-qualified status — so a verifier can independently confirm the qualified status of the timestamping evidence.

The methodology does not claim what the timestamp does not prove. Qualified timestamping establishes that data existed in a particular state at a particular moment. It does not establish the commercial truth of the underlying claim. That separation is preserved in every Keigen verifier output.

What the Keigen Evidence Methodology is not

Three categorical exclusions matter for accurate scope.

The methodology is not a qualified trust service. Keigen does not currently hold UK or EU qualified trust service provider status. The methodology integrates with qualified timestamps issued by listed providers but does not provide qualified trust services itself.

The methodology is not legal advice or forensic certification. Evidence packs produced under the methodology are decision-useful for commercial and early compliance contexts. They are not forensic evidence under ISO/IEC 27037 certification, and they do not constitute legal opinion. Where legal proceedings or formal forensic certification are anticipated, Keigen recommends engagement with appropriate legal counsel and accredited forensic specialists alongside the evidence pack.

The methodology is not employee surveillance, productivity monitoring, or worker tracking. The work-evidence pillar (Pillar 4) is explicit on this distinction: the question is what happened to the work, not what happened to the worker. Keigen products in this pillar verify deliverable evidence, AI-agent activity records, and approval chains — not keystrokes, screen captures, or worker-level activity trails.

The methodology is not “evidence for everything.” The graduation principle is a hard operational boundary. Decisions verifiable through normal business action receive Level 1 outputs. Heavy evidence infrastructure is reserved for decisions where downstream external review is predictable. Applying Level 3 machinery universally would degrade the methodology’s commercial usefulness.

How to apply the methodology in your operation

Three diagnostic questions identify whether evidence infrastructure is the right intervention for a given commercial decision.

First: who reviews this decision later, and on what authority? If the answer is “the same people who made it” or “no one with adversarial standing,” the decision likely belongs at Level 1. If the answer involves a sponsor’s CFO, an auditor, a regulator, an insurer, a payment provider, a counterparty, or a public stakeholder, the decision likely sits at Level 2 or Level 3.

Second: what would a successful challenge to this decision look like? Walk through the form a challenge would take. If the challenger asks “show me what happened, who acted, under what rules, and what evidence can support review” — and the answer would be a screenshot, an email thread, or someone’s recollection — the decision is exposed and would benefit from evidence infrastructure. If the answer is a structured artifact that a competent reviewer could interrogate, the decision is already adequately supported.

Third: which of the six pillars does this decision sit within? B2B website and pre-form sales decisions sit within Pillar 2 (BuyerRecon). Ecommerce campaign and growth-quality decisions sit within Pillar 3 (RealBuyerGrowth). AI-agent and offshore work-acceptance decisions sit within Pillar 4 (TimeToPoint). Allocation, prize-draw, and sponsor-funded selection decisions sit within Pillar 5 (Fidcern Verified Draw). Decisions outside these four pillars typically require a custom-scoped engagement rather than an off-the-shelf product.

Together, these answers determine the level of evidence infrastructure that fits.

Which methodology fits your operation?

Keigen Evidence Methodology FAQ

What is the Keigen Evidence Methodology?
The operational discipline of producing evidence designed to support external review, applied selectively to commercial decisions where the stakes warrant it. It operates through the Attention Monetary System (AMS) framework’s five layers — Intent, Attention, Trust, Policy, Governance — graduating from Level 1 (verifiable through normal business action) through Level 3 (full evidence pack with qualified timestamping).

Is the methodology the same as the AMS framework?
The Keigen Evidence Methodology is the human-readable name for the discipline. The Attention Monetary System (AMS) framework is the technical architecture the methodology operates through. The two describe the same body of work at different levels of abstraction.

Does every Keigen product produce a full evidence pack?
No. Evidence infrastructure is graduated. BuyerRecon produces Level 1 outputs by default; RealBuyerGrowth produces Level 1 with a Level 2 add-on for high-ticket and high-dispute paths; TimeToPoint produces Level 3 for AI-assisted work; Fidcern Verified Draw produces Level 3 as core feature.

How does the methodology relate to compliance frameworks like the EU AI Act, ISO/IEC 42001, or NIST AI RMF?
The methodology integrates with these frameworks rather than replacing them. Evidence produced under the methodology is structured to support compliance filings against the relevant standard. The 2027 bridge thesis (Pillar 6) is that methods built in 2026 for commercial defensibility become 2027 compliance evidence as enforcement dates take effect.

Is Keigen a qualified trust service provider?
No. Keigen is not a UK or EU qualified trust service provider. Qualified electronic timestamps used in Level 3 evidence packs come from EU-listed providers through API integration. Keigen records the trusted-list status of every timestamp at issuance time so verifiers can independently confirm qualification.

Does the methodology work alongside our existing tools (CRM, analytics, AI platforms, attribution)?
Yes. The methodology reads from existing systems rather than replacing them. Each Keigen product is built to operate alongside the analytics, CRM, AI infrastructure, and identification tools already in place — adding the evidence layer above them rather than competing with them at the data layer.

What does “verifiable through normal business action” mean?
It means the buyer of an evidence output can confirm or challenge the verdict by performing routine work — a phone call, a CRM check, a manager review, a delivery confirmation — without needing the evidence pack’s contents. Where verification requires structured artifacts the buyer cannot independently reconstruct, the decision is at Level 2 or Level 3 territory.