Sales records
More leads marked “qualified.”
Read sample record: Sales records
Enquiries are rising, and more prospects meet our qualification criteria. The number of sales meetings has stayed flat.
Company intelligence for people and AI
Put what your company learns to work.
We’re building sho to develop a working understanding of your business from its records, revise it as evidence changes, and bring it into the work of people and AI.
In development. Exploring first implementations.
From events to context
An assistant asks for context. MCP carries back a working belief, its sources, and uncertainty.
One connected systemIllustrative
About this illustration
- Events from your sources. Changes arrive from documents, conversations, and approved tools.
- One normalizer. Different events become consistent records. Their sources stay attached.
- A comprehension graph. Records connect as they arrive, revealing the work they inform.
- An epistemic overlay. Connected records support a tentative explanation. Its reasons and uncertainty stay attached.
- Context through MCP. An assistant asks for context. MCP carries back a working belief, its sources, and uncertainty.
The comprehension graph connects source records. Here, reply history shows slower responses, and loss notes describe prospects leaving before a reply. Together, they support a tentative explanation: slow replies may be costing sales. The epistemic overlay keeps that working belief with its reasons, alternatives, and uncertainty. A planning assistant requests context through MCP and receives the belief with its supporting sources. Follow the example below to see how it could shape a plan, change with new evidence, and inform another assistant’s work. This illustrates intended behavior, not a live product.
A worked example
How sho can help.
A software business has more leads, but no more sales meetings. Follow how sho could connect the records, help the team test an explanation, and inform the next reply. Illustrative walkthrough using sample records and conversations.
sho develops a belief from the company’s own records.
Reply history
The team is taking longer to respond.
Read sample record: Reply history
First replies to new enquiries are taking longer than in the previous period.
Loss notes
Some prospects leave before a reply.
Read sample record: Loss notes
Some lost prospects say they chose another provider before our team responded.
sho connects these observations around the same prospects, replies, and outcomes.
Working belief Tentative
“We may be losing prospects because we reply too slowly.”
An explanation drawn from the records. No single record says this.
- Why it may fit
- Replies are slower. Some prospects leave before a response.
- Another possibility
- The leads may be a poor fit for the product.
- What would weaken it
- Faster replies, but no higher share booking meetings.
sho helps the assistant build a better plan.
The working belief changes how the team approaches growth: test where prospects are being lost before spending more to attract them.
Planning assistantContext from shosho updates its view when evidence changes.
Later, the team has new evidence
- The trial result
- Replies are faster. The share booking meetings stays flat.
- Another clue
- Some “qualified” leads are solo users. The product is built for teams.
The first explanation weakens. Its evidence and reasoning remain available.
Revised working belief Still tentative
“We may be attracting solo users to a product built for teams.”
Product fit becomes a stronger possibility. The cause is still uncertain.
sho carries that learning into the next assistant.
A different assistant uses the revised view to shape a new piece of work. It checks team size before suggesting a meeting.
The belief travels with its evidence, uncertainty, and revision history.
Email assistantContext from sho
Across the business
Company learning, across the business.
A lesson from one part of the business can change a decision in another. sho is being built to carry that understanding across people, teams, and AI tools.
Understand what may be happening.
Bring sales notes, customer feedback, and delivery updates together to examine why work is slowing down or an opportunity is being lost.
Decide what to do next.
Prepare a launch brief, a customer reply, or a follow-up plan with the relevant commitments, owners, and unresolved questions.
Let the next task benefit.
When a decision changes or an explanation weakens, give the next brief, draft, or handoff the updated context and its reasons.
How it’s designed to work
From company records to working understanding.
Connect the evidence, develop an explanation, and keep that view open to revision.
Connect the records.
Organize records around people, projects, decisions, and commitments. These shared definitions and relationships form the comprehension graph, with sources and access rules attached.
Develop a working view.
Develop explanations from connected records. The epistemic overlay keeps each working belief tied to its reasons, alternatives, and uncertainty, distinct from accepted decisions. As evidence changes, sho revises the company view.
Bring it into the work.
Give people and agents the company understanding relevant to their task and permissions. sho’s agent runtime uses it to carry out work. MCP is the planned connection for compatible AI tools.
Agent runtime
Give the work to an agent.
sho’s planned agent workspace gives each agent the company context and approved tools for its task. Ask it to prepare a brief, implement and test a code change, or update a task after approval. Follow progress and review the result.
A working view can change.
Its reasons should stay clear.
Keep inferred explanations distinct from accepted decisions, with the evidence and history behind both.
TentativeA working explanation, open to revision.
Keep the reasons, alternatives, and uncertainty visible, including evidence that would weaken the explanation. An inferred belief can inform a task without becoming an accepted decision.
AcceptedA decision with someone accountable for it.
Accepted by the person responsible, with supporting evidence and a clear scope. Acceptance does not make every explanation behind a decision a proven fact.
DisputedDifferent sources, different answers.
Keep each position and its sources visible. Record any resolution, who made it, and why.
StaleSomething it depends on has changed.
Flag information that needs checking because time has passed, or its source or circumstances changed. Keep replaced decisions in the history, out of current instructions.
UnknownThere isn’t enough available evidence.
Say what is missing. Information that a person cannot access is not included in their context.
Control and ownership
You decide what
sho can use.
These are the boundaries a first implementation needs to prove.
- You choose the access.
- Start with approved sources. People and AI should see only the information they’re allowed to use.
- Actions need their own approval.
- Connecting a tool is permission to read, not to change it. Actions require a named person’s approval and a checked result.
- The sources stay in view.
- Working beliefs and decisions keep their supporting evidence and revision history. Missing information, uncertainty, and disagreements remain visible.
- Your company knowledge stays yours.
- Your company’s information and its history would be exportable. We agree how long information is kept and how it can be deleted before launch.
A few practical questions
Before we talk.
What can we do with sho today?
sho is in development. We’re building and testing the path from company records to useful agent work. For a first implementation, we agree a useful task and its sources, then test the result and how it changes when the evidence changes. The walkthrough shows intended behavior using sample records.
We already use AI. What would sho add?
AI assistants and search tools can already help you find information and draft work. sho is being built to develop and maintain a view of company decisions, explanations, and open questions as evidence changes. Our planned agent workspace would use that understanding to carry out tasks. Compatible AI tools could request relevant context through MCP, the Model Context Protocol, under your access rules.
Does sho replace our existing tools?
Your documents, conversations, and business systems stay where they are. sho is designed to connect their evidence and develop a working understanding of the business, while preserving sources and access rules.
Can sho make changes on its own?
Reading, proposing, and acting are separate permissions. The design requires a named person’s approval before a change to a real system, followed by a check of what actually happened.
Where would it run, and what could we export?
We would agree where sho runs based on your security and operating requirements. Your company’s information, sources, history, and work would be exportable; sho keeps its software. We agree where data is kept, how it is backed up, and when it is deleted before connecting your sources.
Start with your business
Start with the work you want to improve.
A decision that takes too much digging. A handoff that loses context. A recurring task that starts from scratch. Let’s work through one with your team and define what a better result would look like.