Company Intelligence · Roadmap engagement · Start scoped

Give the important parts of your business an intelligence layer

Turn scattered company knowledge into persistent, usable intelligence for the people responsible for projects, products, customers, teams and services.

Send us one business area See the roadmap

Your information exists. Your intelligence is fragmented.

What a scoped intelligence layer does
rememberretrieveobserveconnectchallengecreaterecommendoptionally act
What the human stays
accountablejudgment-bearingdecision-makingrelationship-owninglegally responsible
01 · The problem

Your information exists. Your intelligence is fragmented.

Your company probably does not lack information. It already exists across CRM, ERP, documents, project tools, email, Slack or Teams, analytics, meeting notes, support tickets, research, contracts, dashboards and people's heads. The problem is that the right context is rarely available at the right moment.

Projects forget why decisions were made. Sales teams lose customer history. Product teams rediscover old insights. Consultants repeat work already solved elsewhere. Experienced employees become critical because they simply remember more than everyone else.

Most companies have already built systems of record. CRMs remember contacts and deals. ERPs remember transactions. Project tools remember tasks. Analytics platforms remember events. Document systems remember files. But important business work depends on more than records.

People need to knowWhy was this decision made?
People need to knowWhich assumptions supported it?
People need to knowWhat changed since then?
People need to knowWhich source is authoritative?
People need to knowWhat have we already tried?
People need to knowWhich customer commitments must not be forgotten?
People need to knowWhich signal deserves attention now?
People need to knowWhat should the responsible person consider before acting?

Today, answering these questions often requires reconstructing context manually across multiple systems and people. That creates an organizational intelligence gap.

02 · The cost

Rarely one line in a budget. Always friction across the company.

Repeated work

Teams rediscover information, rebuild briefs, repeat analysis and revisit questions the company has already answered.

Slower decisions

Important meetings begin with context reconstruction instead of decision-making.

Lost organizational memory

When experienced employees leave or change roles, part of the company's operating context leaves with them.

Weak handovers

Projects, accounts and services lose continuity when ownership changes.

Inconsistent customer experience

Customers repeat themselves because one part of the organization does not fully remember what another part already learned.

Missed signals

Product, commercial and operational teams see individual pieces of evidence but fail to connect them early enough.

Underused expertise

Expert knowledge helps the immediate situation but often fails to become reusable organizational capability.

The result is not simply wasted time. It is a company making important decisions with incomplete memory.
03 · The idea

Scoped Intelligence

Persistent AI-enabled intelligence attached to a defined business object or area of responsibility. The object might be a strategic project, a product, a customer or account, a service line, a team, a GTM motion, a market, a contract, a deal, an asset, a portfolio, or another important part of the business.

The space this occupies

This is the space between a generic AI assistant and full automation. The objective is not to make AI own everything. The objective is to make sure important work no longer operates without memory, context and intelligence.

04 · In practice

What this looks like on real business objects

Five patterns we see most often. Each attaches to something the business already treats as important.

Pattern 01

Project Intelligence

scopeobjectivesstakeholdersdecisionsassumptionsdependenciesriskslessonscurrent status

What changed this week?

Which assumptions remain unvalidated?

Which decision contradicts the original requirements?

What did the client actually approve?

What should the steering committee worry about tomorrow?

This is not necessarily AI managing the project. It is a project that can explain itself.

Pattern 02

Customer / Account Intelligence

relationship historystakeholderscontractscommitmentsobjectionssupport issuesproduct usagecommercial exceptionsopportunitiesmarket signals

Before a meeting: tell me what I cannot afford to forget about this customer.

After a meeting: what changed?

Six months later: why did we agree to this exception?

The salesperson does not disappear. The salesperson becomes the person in the room who remembers everything relevant.

Pattern 03

Product Intelligence

researchcustomer interviewsusage datasupport ticketsroadmap decisionsexperimentscompetitorspricingdesign decisionstechnical constraints

What evidence supports this roadmap item?

Which customer problem is becoming more important?

Have we already tried this?

Which support issues contradict the current product strategy?

The goal is not an AI product manager. The goal is a product team that can reason with more of its accumulated evidence.

Pattern 04

GTM Intelligence

ICP assumptionsaccount datasales callswinslossesobjectionscampaignscontentpricingcompetitorspipelinemarket signals

What have we actually learned about this market?

Which ICP assumptions are no longer supported by our wins?

What patterns appear across the last 30 lost deals?

What should sales know before contacting this account?

AI can write an email in seconds. The intelligence behind the email is the advantage.

Pattern 05

Service Intelligence

delivery standardstemplatesbenchmarksrecurring failure modescustomer exceptionsspecialist knowledgesuccessful approacheslessons from previous engagements

For consultancies, agencies, software-development firms, engineering businesses and other professional-services companies, every engagement creates knowledge. Client 47 should not start from zero. They should benefit from what the organization learned while serving Clients 1–46.

That is how expertise begins to compound.

05 · The architecture

A useful Intelligence Layer is more than a chatbot connected to files.

Seven elements need to work together.

01

Scope

What exactly should this intelligence understand? A strategic account? A product? A transformation program? One GTM market? Clear scope improves relevance, permissions, ownership and evaluation.

02

Memory

What should survive meetings, employee changes and AI sessions? Decisions, assumptions, lessons, exceptions, definitions, previous outcomes, rationale.

03

Live signals

What must remain current? CRM, ERP, analytics, project state, tickets, product usage, pipeline, market data.

04

Skills

What should the intelligence know how to do? Prepare a risk review, analyze a lost deal, build an account briefing, check scope creep, prepare a steering-committee update, compare customer signals.

05

Reasoning

How should AI combine what happened, what we know, what matters and what usually works — and turn it into useful guidance now?

06

Interface

Where should people use it? CRM, Slack or Teams, project software, a dedicated workspace, dashboards, meeting preparation, existing internal tools.

07

Authority

What may the intelligence do — and where does that stop? These are different levels, not one switch.

Element 07 · the authority ladder
knowretrieveanalyzerecommenddraftmodifyexecuteapprove
Capability does not equal authority

AI may identify contractual risk without being allowed to accept contractual risk. It may recommend pricing without being allowed to commit pricing to a customer.

This is the same distinction that governs agents in production — a technical permission is not a business mandate. We treat it in full in the Agent Owner's Manual.

06 · Sequencing

Why we do not start with one giant "company brain".

The goal is not to connect everything first. The better starting point is a valuable, bounded business object — Project Apollo, top 30 strategic accounts, DACH enterprise GTM, one product line, one service line.

A bounded scope makes it easier to define what matters, what does not, who owns it, which sources are authoritative, which permissions are required, what the AI may do, and how usefulness will be measured.

Start scoped. Prove value. Then connect what works.
07 · The stack

You do not need to replace your current systems.

Different layers play different roles. The intelligence layer sits between them rather than on top of them.

Figure 1 · The Company Intelligence stack

Five layers, one accountable participant

Durable semantic context
Documents and structured files
Policies, decisions, assumptions, instructions, lessons, procedures, definitions and skills.
Live operational state
CRM, ERP, databases and operational tools
Current records, transactions, project state, usage, pipeline, telemetry and other live signals.
Reasoning layer
AI
Connect history, live state, context, goals and skills.
Access and action
APIs, MCP, connectors and tools
Retrieve live information and perform approved actions.
Accountable participant
Human
Own important decisions, relationships, approvals, trade-offs and responsibility.
This is why simple structured artifacts — including human-readable Markdown — can be useful alongside operational systems. Markdown is not the database; it can be part of the semantic context layer that helps both humans and AI understand the business object.
08 · Sequencing

Intelligence before autonomy

Company Intelligence does not require jumping directly to autonomous agents. A useful maturity model has seven levels — and most of the value is not at the top.

Figure 2 · The intelligence maturity model

Many organizations create significant value at Levels 2–4

Level 0
InformationWe store things
Level 1
SearchWe can find things
Level 2
KnowledgeWe organise and explain things
Level 3
IntelligenceWe connect knowledge, history, signals and goals
Level 4
AssistanceAI helps humans decide and create
Level 5
AgencyAI performs approved actions
Level 6
AutonomyAI owns bounded outcomes
Highlighted levels are where most organizations find value first. The objective should not automatically be maximum autonomy — it should be the right level of intelligence and authority for the work.
09 · What we build

The Company Intelligence Roadmap

We help companies move from broad AI ambition to a concrete Company Intelligence architecture. The first substantial engagement is typically the roadmap.

The roadmap answers: which parts of the business are strong Intelligence Object candidates? Where is context currently fragmented? What should the company stop forgetting? Which decisions could improve with stronger intelligence? Which systems and knowledge sources are required? Which information should become persistent memory? Which live signals matter? Which reusable skills are needed? Where should employees interact with the intelligence? Who remains accountable? What authority should AI have? Which opportunities are valuable enough to implement first?

What it is not

The goal is not to produce a long list of disconnected AI use cases. The goal is to create a coherent map of what should become intelligent, why, and in what sequence.

The deliverable

01

Intelligence Object Map

A view of candidate objects across product, customers, GTM, delivery, operations, projects, services, knowledge and management.

02

Priority Matrix

Each candidate assessed against business value, context fragmentation, information availability, repeated decision frequency, risk, ownership and implementation complexity.

03

Intelligence Canvas

  • business purpose
  • human owner
  • users
  • decisions supported
  • key questions
  • persistent memory
  • live signals
  • sources of truth
  • required skills
  • interface
  • authority
  • evidence rules
  • freshness requirements
  • evaluation criteria
04

Implementation Sequence

A clear Now → Next → Later recommendation, including the best candidate for the first working Intelligence Layer.

05

First Pilot Definition

  • intended outcome
  • required systems
  • context and memory requirements
  • user group
  • authority level
  • success criteria

What happens after the roadmap

The roadmap has standalone value. Your organization can use it internally, with your existing technology partners, or with auxfirst. If useful, auxfirst can also help implement the first Intelligence Layer — context architecture, persistent memory, source-of-truth design, RAG and retrieval, live system connections, reusable AI skills, user interaction design, authority and approval rules, evidence and freshness standards, evaluation, and feedback and learning loops.

We recommend starting with one important bounded object rather than trying to redesign the whole company at once: one strategic project, one product, top strategic accounts, one GTM segment, or one service line.

10 · Fit

What a good first project looks like

Strong first candidateWeak candidate
High business importanceNobody owns the underlying business object
Repeated decisionsSource information is unreliable
Significant existing informationThe business impact is trivial
Fragmented contextThe workflow is too rare to justify persistent intelligence
Several stakeholdersThe idea exists mainly because a chatbot would look impressive
Identifiable human ownership
Frequent need for preparation or analysis
Evidence that can be checked
Clear potential users

The objective is not to deploy AI somewhere. The objective is to improve an important part of the business.

A simple diagnostic you can run today

Pick one important project, customer, product, service or team. Ask: what should we never have to rediscover about this object? Then ask:

If the answers are fragmented, you may not primarily have an AI problem. You have an organizational intelligence gap.

11 · The engagement

Company Intelligence Roadmap

A focused engagement designed to turn the broad idea of "using AI better" into a prioritized architecture for your business.

You receive

Eight artifacts

  • Company Intelligence opportunity map
  • Prioritized Intelligence Objects
  • Intelligence Canvas for priority areas
  • Source and memory requirements
  • Authority and human-accountability model
  • Implementation priorities
  • First-pilot recommendation
  • Clear Now / Next / Later roadmap
Working format

Designed with your stakeholders

Built with the relevant business and technology stakeholders. Can be preceded by an online or onsite Company Intelligence briefing where leadership alignment is useful.

Investment

Fixed fee

Agreed before the work begins, based on the scope and number of business areas included.

The outcome

A concrete answer to: what should become intelligent in our company first, what will it require, and how should we implement it without creating another disconnected AI experiment?

Why start with the roadmap

A platform decision does not answer what should become intelligent, what should be remembered, which sources should be trusted, which questions matter, which skills are required, who owns the outcome, how much authority AI should receive, or how success should be measured. Those are business-design questions. The roadmap connects them before implementation begins.

It helps avoid two common outcomes:

Failure mode 01

A collection of AI experiments

Many isolated prompts, automations and assistants with no shared architecture.

Failure mode 02

An oversized AI transformation

A complex program trying to connect everything before proving value anywhere.

Start scoped. Build useful intelligence. Scale from evidence.
12 · The bigger opportunity

We spent decades building systems of record.

The next opportunity is building systems that can also help us understand meaning. Why was this decision made? What assumption supported it? What changed? What contradicts it? What should we pay attention to now?

The future of enterprise AI is not only AI doing more work. It is every important part of the business becoming more intelligent: remembering its history, understanding its current state, bringing relevant expertise into the moment, helping humans make better decisions.

The company of the AI era may not be the company where agents do everything. It may be the company where nothing important has to operate without memory, context or intelligence.
Questions

Before you ask

Do we have to connect everything first?

No. The goal is not to connect everything first. The better starting point is a valuable, bounded business object — one project, one product, top strategic accounts, one GTM segment or one service line. Start scoped. Prove value. Then connect what works.

Does this replace our CRM or ERP?

No. You do not need to replace your current systems. Different layers play different roles: documents hold durable semantic context, CRM and ERP hold live operational state, AI is the reasoning layer, APIs and connectors are the access and action layer, and the human remains the accountable participant.

Do we need autonomous agents for this?

No. Company Intelligence does not require jumping directly to autonomous agents. Many organizations can create significant value at Levels 2 to 4 of the maturity model. The objective should not automatically be maximum autonomy — it should be the right level of intelligence and authority for the work.

What does the roadmap cost?

Fixed-fee engagement, agreed before the work begins based on the scope and number of business areas included.

Next step

Send us one business area you believe is a strong candidate.

Strategic accounts, one product, one major project, one GTM motion or one service line. We can use that as the starting point for a Company Intelligence discussion and determine whether a Roadmap engagement is useful.

Start the conversation See the Agent Owner's Manual

Designing how businesses and people work with increasingly capable AI.

Where this sits in the auxfirst canon

Closest engagements
Process Brains — turning a workflow into a defensible agentic service The Agent Owner's Manual — the operating standard for agents once intelligence becomes action The Agent Operability Audit — whether an agent can actually work one specific workflow Agent Process Design — the practice behind the engagement
Free frameworks that go deeper
Five Steps to Becoming an Agentic Organization — the pillar playbook Crossing the Chasm to the Agentic Organization — why pilots stall between demo and adoption The Agent-Operable Enterprise — the field guide, with a scored 42-item self-assessment The 10 AUX Heuristics — the working principles
On the context layer
MD files explained — the format the semantic context layer is written in AGENTS.md for teams that don't write code — and why a short context file beats a complete one Prompt injection in .md context files — the trust boundary around instruction files
Reference
The auxfirst knowledge base — every AUX concept, defined and cross-linked What Is Agentic User Experience (AUX)? — the discipline underneath AI info page — how machines should read auxfirst

Method · Scoped Intelligence, the Intelligence Canvas and the maturity model described here are auxfirst's own frameworks, developed through client work rather than derived from a published standard — we say so plainly rather than implying one exists. The examples are illustrative patterns, not case studies; auxfirst does not publish named client results. Pricing is stated exactly as it is offered: a fixed fee agreed before work begins, scoped to the number of business areas included.

Emil Krzemiński is the founder of auxfirst, the agentic experience design agency — helping product, developer and business teams design AI systems that remember, adapt, and earn the right to act. Start with the Agent Owner's Manual, the pillar playbook Five Steps to Becoming an Agentic Organization, or a conversation. For how machines read auxfirst, see the AI info page. Subscribe to the auxfirst Substack for what's next.