BloomSpec

Employee Knowledge & Action

The problem of organisations accumulating knowledge across dozens of systems, where an employee cannot reliably find it, trust it, or act on what it reveals. Scoped by who retrieves, not who benefits: an employee retrieving is in scope regardless of who eventually reads the output; a customer or an agent retrieving on a customer's behalf is not. Action is in scope only when it is knowledge-derived — it could not have been requested because nobody knew there was anything to request until synthesis revealed it. Not AI-scoped: the domain is organisations losing, mistrusting, and failing to act on their own knowledge, which predates AI by decades.

208 pages

Agent patterns16

Bounded Task Execution Agent
For low-risk, internal-only actions, requiring human review on every single instance is itself a bottleneck research found production deployments now avoid for…
Byproduct Capture Agent
KCS's own core insight: knowledge captured as a byproduct of solving a problem is more complete and more likely to happen than knowledge authored separately…
Completion Verification Agent
Manual spot-checking of every completion does not scale to fulfilment-centre volumes; automated verification with a human override path does.
Connector Onboarding Agent
Manually configuring each new connector is the single largest driver of onboarding time for a new tenant or a newly acquired subsidiary.
Content Verification Agent
Drift between scheduled reviews (CAP-113) is invisible until the next cycle catches it; a continuously running rule set closes that gap without waiting for a…
Contradiction Resolution Agent
The action could not have been requested because nobody knew there was anything to request until the contradiction itself was found — the domain's own origin…
All 16 agent patterns →

AI opportunities51

Activity-graph expertise inference
Ranking likely subject-matter experts per topic from authored content, resolved tickets, and code contributions.
Adaptive interview question generation
Conducting a conversational interview that generates follow-up questions based on the expert's prior answers.
AI-assisted content drafting
Generating a first draft, summary, or refinement of a knowledge card or article from a prompt or source material.
AI-assisted redaction drafting
Proposing redactions of client-identifying detail in engagement content, for human review before promotion.
Auto-archive threshold detection
Flagging content as an archive candidate once its usage has decayed below a threshold.
Auto-draft generation from resolved interaction
Summarising a resolved ticket or thread into a structured candidate knowledge article.
All 51 AI opportunities →

Capabilities46

Action Audit Trail & Attribution
Recording, for every agent action, what was done, on whose authority, under which policy, and why — retained for accountability the same way an audit trail is…
AI-Assisted Procedure Drafting from Expert Input
Generating a first-draft, step-sequenced procedure from a captured expert interview or a raw process description — the bridge from elicited tacit knowledge…
Answer Confidence & Grounding Verification
Checking that a generated answer is genuinely traceable to the passages it cites, rather than accepting a plausible-sounding synthesis at face value — the…
Automated Content Quality Rules
Continuous, algorithmic verification and un-verification of owned content based on behavioural, content, and analytical signals — running nightly rather than…
Bounded Autonomous Execution
An agent actually carrying out the recognised action — editing a document, updating a ticket, posting to a project tool — within an explicitly defined autonomy…
Byproduct Capture During Resolution
Automatically drafting a candidate knowledge article from a resolved ticket, answered question, or closed incident, without requiring the person who solved it…
All 46 capabilities →

Entities22

Access Partition
Defines a partition boundary finer than tenant level — an engagement, a matter, a project — within which content and authorship are scoped, independent of role…
Agent Access Finding
The logged record of a single Agent Policy evaluation — which policy fired, what action it was evaluated against, and what it resolved to.
Agent Action
A proposed or executed action taken by an AI agent on the organisation's behalf, carrying its origin (synthesis-triggered, per the domain's scope test), the…
Agent Policy
A governance rule evaluated at the moment an agent attempts a tool call, resolving to block, filter, or flag for review.
Chunk
A retrievable fragment derived from a Source Document or Knowledge Item, carrying its own inherited permission scope and embedding.
Completion Record
Evidence that a specific person completed a specific Procedure, at a specific version, including any required photo or comprehension-check evidence.
All 22 entities →

Glossary20

Agent access policy
A governance rule evaluated at the moment an AI agent attempts to call a tool, resolving to block, filter, or flag the action for human review.
Chunk
A fragment of a document or knowledge item, small enough to embed and retrieve individually, rather than searching whole documents.
Command-routed action
An action a person explicitly asked for — 'create a ticket for this' — which any workflow automation tool can already do, and which this domain deliberately…
Content verification
The act of a person or an automated rule confirming that a piece of knowledge is still accurate, distinct from the general software-testing sense of…
Coverage disclosure
A system's explicit statement of what it does and does not have material to answer, rather than answering every question with the same apparent confidence.
Cross-source contradiction
A case where two currently published, live sources disagree about the same fact — as opposed to one source being an outdated version of the other.
All 20 glossary →

KPIs14

Byproduct capture rate
Share of resolved incidents or tickets that produce a confirmed Runbook Entry.
Contradiction resolution time
Time from a contradiction being flagged to it being resolved or dismissed by an accountable owner.
Cross-boundary promotion rate
Share of closed engagements or projects that produce at least one piece of sanitised content promoted to the firm-wide playbook.
Cross-source contradiction detection rate
Share of genuine, live cross-source contradictions that are actually flagged before an answer is generated from disagreeing sources.
Duplicate-match resolution rate
Share of new incidents correctly matched to an existing Runbook Entry rather than re-diagnosed from scratch.
Ethical wall exception rate
Rate of attempted cross-partition access, whether blocked, flagged, or genuinely exceptioned via a reviewed waiver.
All 14 KPIs →

Personas9

Engagement Consultant
Delivering client engagements, drawing on and contributing to the firm's knowledge base within the bounds of whichever engagement he is staffed to.
Field Service Technician
Diagnosing and resolving equipment faults on customer sites, using whatever documentation exists to do it right the first time.
Fulfilment Centre Worker
Operating equipment and following safety procedures correctly during a shift, often with limited prior training.
Go-to-Market Team Member
Building customer-facing proposals and responses from internal engineering and product knowledge.
Knowledge and Methods Lead
Deciding what closed-engagement learning can be sanitised and promoted into the firm-wide playbook, and owning the platform's governance configuration more…
Knowledge Curator
Content verification, review-cycle management, and support-ticket triage for the shared knowledge platform.
All 9 personas →

Products4

Glean
The unified platform — search, assistant, and Agent Library with its Protect governance layer — sold as a single product rather than a portfolio of separate…
Guru
The unified knowledge platform — verified content, Knowledge Agents, and MCP-based external access — sold as a single product.
Notion AI
Enterprise Search across the workspace and connected tools, with Custom Agents and an external agents API exposing the workspace to third-party AI systems.
Rovo
Search, Chat and Agents over the Teamwork Graph, sold as a layer across the Atlassian suite and connected third-party apps rather than as a standalone product.

Scenarios7

A field technician finds the right bulletin before a customer visit
A Calder Field Service technician needs to interpret an unfamiliar fault code with fifteen minutes before departure. Deliberately the happy path for Discovery…
A finding becomes a correction, inside the gates
Directly continuing SCN-103: the flagged milestone contradiction is tagged action-worthy, and a proposed correction moves through drafting, tool-call…
A wall holds, and a lesson gets sanitised across it
A Kestrel consultant drafts client-specific content inside an ethical wall; a consultant on a competing client's engagement is correctly shown nothing; months…
One step at a time, on the floor
A Haddon fulfilment-centre worker operates equipment requiring a documented safety procedure, delivered and verified step by step rather than as a document to…
The fix nobody wrote down, until this time
A Marrow on-call engineer resolves a 3am outage; the resolution becomes a structured runbook as a byproduct rather than separate documentation work; a later,…
The plan says one date, the report says another
Marrow's GTM team searches for a ship date and receives two disagreeing live sources, worked through end to end — the domain's own canonical example from its…
All 7 scenarios →

Solution areas7

Content Authoring & Lifecycle
Deliberately creating, owning, reviewing, and retiring knowledge, with an accountable person behind each piece of content. Includes ethical-wall and…
Expertise & Tacit Knowledge
Capturing what was never written down, by elicitation rather than indexing: structured interviews, cross-person contradiction flagging, locating who knows…
Knowledge-Derived Action & Agent Governance
Action that could not have been requested because nobody knew there was anything to request until synthesis revealed it — bounded explicitly against…
Knowledge Discovery & Retrieval
Connecting to the fragmented source landscape, ranking, grounding, and answering. Includes permission-aware retrieval (index-time and query-time enforcement,…
Structured Process & Procedural Knowledge
Standard operating procedures, work instructions, and other step-sequenced, safety- or compliance-critical content, with requirements (versioned procedure…
Trust, Quality & Detection
Algorithmic assessment of what already exists: staleness, cross-source contradiction between two live documents, source-quality and confidence signals.…
All 7 solution areas →

Vendors8

Atlassian
Work-management vendor whose Rovo layer adds enterprise search, conversational AI and agents on top of Jira and Confluence, grounded in the Teamwork Graph…
Coveo
Relevance and search platform selling a single underlying engine into both customer-facing and employee-facing deployments.
Glean
Enterprise AI search and agent platform positioned as connecting and reasoning across an organisation's work applications. Core competence is broad-connector…
Guru
Knowledge management platform positioned around verified, governed content, extended with agentic retrieval across connected sources and an MCP server exposing…
KNOA
AI-led structured interview platform for capturing tacit knowledge ahead of employee departure, notable for cross-referencing multiple respondents' answers on…
Microsoft
Microsoft 365 Copilot, the AI assistant layer across the Microsoft 365 suite, grounding responses in SharePoint, Exchange, and Teams content subject to…
All 8 vendors →

Verticals4

Consumer Platforms and Fulfilment
Multi-sided platform businesses with a large frontline operational workforce and an external partner or seller network.
Industrial Manufacturing
Equipment manufacturers where engineering judgement, safety procedures, and field-service knowledge are the dominant knowledge assets.
Professional and Consulting Services
Firms delivering client engagements, where confidentiality obligations partition knowledge within the firm itself.
Software and Technology
Engineering-driven companies where technical documentation is tied to a fast release cadence and institutional memory is largely incident-driven.