05 / Products
Agentic systems for real organisational work.
Linear Horizon Platform is an operating system for AI. Specialist products then apply that system to analysis, research, strategy, growth and customer service.
Figure · Product system
Operating system
01
Platform
An operating system for AI.
AI operating system
Agents, tools, memory, permissions and controls that let AI operate around the work your organisation already does.
Applications
02
Stack
Agents that analyse.
Analysis/Agentic Data Analysis
Ask open-ended questions of your data.
An agentic data analysis system that investigates open-ended questions, tests hypotheses and produces a structured analytical deliverable.
03
Solve
Agents that reason.
Research/Agentic Hard-Problem Research
Start a run. The tree is the ledger.
An unattended problem solver: each tree node is axioms, reasoning and an evaluating agent. A run wakes high-scoring or interesting nodes to ideate, Horn-generate or GA-mix.
04
Wargame
Agents that strategise.
Strategy/Agentic Strategy
A machine that can predict the future.
An agentic strategy system that models actors, incentives and possible responses, then simulates how they interact — a machine that can predict the future.
05
Funnel
A local agentic CRM.
Growth/Agentic CRM
Your data never leaves your desktop.
A desktop agentic CRM that researches organisations, qualifies opportunities and drafts outreach locally. Your data never leaves your desktop.
06
Callcenter
Agents that serve customers.
Customer Service/Agentic Customer Service
One conversation, across every channel.
Resolve customer needs across voice, messaging and email, with intelligent human handoff when judgement matters.
07
Query
Agents that investigate usage.
Analytics/Agentic Product Analytics
Understand what users actually do.
Privacy-conscious product and website analytics, where agents investigate the numbers rather than leaving you to read another dashboard.
08
Chat
Agents you can talk to.
Interface/Agentic Chat Interface
A conversation on top of your agents.
A conversational interface to your agents — streaming answers, visible working and voice, on top of the Linear Horizon Platform.
How the products differ
| — | PlatformAct | StackUnderstand | SolveSolve | WargameDecide | FunnelDiscover | CallcenterServe | QueryMeasure | ChatConverse |
|---|---|---|---|---|---|---|---|---|
| Purpose | An operating system for AI | Investigate your data | Search programmes of attack on a hard claim | Predict the future | Local agentic CRM | Serve customers | Measure product usage | Talk to your agents |
| Input | Events, requests, knowledge and tools | Open-ended questions and organisational data | A conjecture, modelling question or precise statement | Strategic problem and actor model | ICP and market signals | Voice, messaging, email and customer context | Product and website events | A question, an agent and a conversation |
| Agentic process | Orchestration, tools and workflows | Hypothesis generation and testing | Choose, expand, evaluate | Actor modelling and simulation | Discovery, research and qualification | Understand, retrieve, act, resolve or hand off | Instrument, measure and investigate | Route, retrieve, act and stream |
| Output | Agentic workflows and actions | Evidence, analysis, visualisations and presentations | A scored tree of attack programmes | Scenario distributions and strategic options | Evidence-backed opportunities | Resolution, action or context-rich human escalation | Metrics, cohorts and investigated findings | A streamed answer with its working shown |
| Channels | Events, APIs and operational interfaces | Analytical environments and presentations | Research chat, node agents, run controls | Strategic briefs and simulation outputs | Desktop app; local CRM data | Voice, WhatsApp, SMS, email, Teams, Slack, Salesforce | Dashboards, event API and MCP | Web chat, voice and token streaming |
| Human role | Oversight and escalation | Review and interpretation | Start, pause, stop; review scores and what remains open | Assumption review and strategic judgement | Approval and relationship building | Judgement, takeover and the work that still needs a person | Interpretation and prioritisation | The person asking, and taking over when needed |
Platform
Act
- Purpose
- An operating system for AI
- Input
- Events, requests, knowledge and tools
- Agentic process
- Orchestration, tools and workflows
- Output
- Agentic workflows and actions
- Channels
- Events, APIs and operational interfaces
- Human role
- Oversight and escalation
Stack
Understand
- Purpose
- Investigate your data
- Input
- Open-ended questions and organisational data
- Agentic process
- Hypothesis generation and testing
- Output
- Evidence, analysis, visualisations and presentations
- Channels
- Analytical environments and presentations
- Human role
- Review and interpretation
Solve
Solve
- Purpose
- Search programmes of attack on a hard claim
- Input
- A conjecture, modelling question or precise statement
- Agentic process
- Choose, expand, evaluate
- Output
- A scored tree of attack programmes
- Channels
- Research chat, node agents, run controls
- Human role
- Start, pause, stop; review scores and what remains open
Wargame
Decide
- Purpose
- Predict the future
- Input
- Strategic problem and actor model
- Agentic process
- Actor modelling and simulation
- Output
- Scenario distributions and strategic options
- Channels
- Strategic briefs and simulation outputs
- Human role
- Assumption review and strategic judgement
Funnel
Discover
- Purpose
- Local agentic CRM
- Input
- ICP and market signals
- Agentic process
- Discovery, research and qualification
- Output
- Evidence-backed opportunities
- Channels
- Desktop app; local CRM data
- Human role
- Approval and relationship building
Callcenter
Serve
- Purpose
- Serve customers
- Input
- Voice, messaging, email and customer context
- Agentic process
- Understand, retrieve, act, resolve or hand off
- Output
- Resolution, action or context-rich human escalation
- Channels
- Voice, WhatsApp, SMS, email, Teams, Slack, Salesforce
- Human role
- Judgement, takeover and the work that still needs a person
Query
Measure
- Purpose
- Measure product usage
- Input
- Product and website events
- Agentic process
- Instrument, measure and investigate
- Output
- Metrics, cohorts and investigated findings
- Channels
- Dashboards, event API and MCP
- Human role
- Interpretation and prioritisation
Chat
Converse
- Purpose
- Talk to your agents
- Input
- A question, an agent and a conversation
- Agentic process
- Route, retrieve, act and stream
- Output
- A streamed answer with its working shown
- Channels
- Web chat, voice and token streaming
- Human role
- The person asking, and taking over when needed
Compared with general agents
Claude Cowork and Grok Bot hand off work. Linear Horizon operates a process.
They are often compared because all three can research, draft and take action. They are not the same kind of product.
How they are used
Products can be deployed, adapted, or combined with services.
You do not have to choose between a product and an engagement. Platform and the specialist applications can be deployed as they are, adapted to existing systems, extended, or combined with consultancy and training. They exist so that not every piece of work starts from a blank page.