Products · 01 Intelligence Platform
One platform. Four stages. The fourth is the one that matters.
Most of this category is very good at the first two stages and quietly hands the rest back to you. What follows is what each stage does, what it needs from you, what it returns, and where it deliberately stops.
01 · GrayMatter Connect
Connect
Your systems already hold the answer. They just hold it separately.
ERP, CRM, HRMS, finance, spreadsheets, the export someone emails every Monday. GrayMatter reads them together so a signal in one system can be seen by the rest.
What it needs from you
- Read access, or a scheduled export, from two or three systems
- Someone who can explain what the fields actually mean in practice
- Enough history to establish a normal pattern for the question asked
What it gives back
- One reconciled view across systems that previously could not see each other
- A record of what was mapped to what, so nobody has to guess later
- An honest report on data quality, including where it is not good enough
02 · GrayMatter Predict
Predict
A dashboard reports the past. A forecast has an opinion about the future.
One forecasting approach pointed at whichever question matters: demand, revenue, admissions, attrition, cash, capacity. Every forecast carries its reasoning and how confident it is.
What it needs from you
- A clearly stated question with a measurable outcome in the history
- Enough completed cycles for the pattern to be real rather than noise
- Agreement on what accuracy would be good enough to be useful
What it gives back
- A forecast with a stated confidence level and the reasoning behind it
- The drivers the model actually weighted, in plain language
- An explicit statement when the data cannot support a reliable answer
03 · GrayMatter Decide
Decide
A number is not a decision. Someone still has to say what to do about it.
The decision layer weighs the forecast against cost, risk and constraint, and returns a specific recommended action with the reasoning attached, so a human can accept it, change it or reject it.
What it needs from you
- The constraints that make an action possible or impossible
- A rough sense of what being wrong costs in each direction
- Who is accountable for this class of decision
What it gives back
- A specific recommended action, not a range of options to interpret
- What it is worth, what it assumes, and what happens if the forecast is wrong
- A clear path to disagree with it, with the reasoning visible
04 · GrayMatter Workflow
Act
Most insight dies in the gap between the meeting and the task.
Once a person approves, the decision becomes work: the request raised, the team notified, the workflow moved, the outcome tracked. Nothing executes without an approval that is recorded.
What it needs from you
- A named human approver for each class of decision
- The workflow or system where the work actually gets done
- Agreement on what may execute automatically and what may not
What it gives back
- Approved decisions turned into assigned, tracked work
- A recorded trail from signal to approval to action to outcome
- Outcomes fed back, so the next recommendation is better informed
On top of the four stages
Five more capabilities. Same platform, same rules.
These are not separate products. They are the surfaces and tools people use to work with the intelligence, and every one of them runs under the same approval and audit rules as the stages beneath it.
-
Conversational intelligence over your business
GrayMatter Copilot
A conversational layer over the connected business. Ask why a number moved, what is at risk this month, or what the recommendation assumed, and get an answer that cites where it came from.
- Answers cite their data
- Scoped to what the person may see
- No answer invented when the data is thin
Demo coming soon
-
Executive command centre
GrayMatter Insights
The four numbers every executive asks for, plus the one almost nobody measures: capacity released, and how much of it was redeployed. Shown as an operating metric, not a footnote.
- Live position across systems
- Recommendations awaiting approval
- Capacity released and redeployed
Demo coming soon
-
Business intelligence and reporting
GrayMatter Analytics
Business intelligence answers what happened. Analytics does that on the same connected data the rest of the platform uses, so the report and the forecast never disagree about the numbers.
- Reports on the reconciled data
- Same definitions as the forecast
- Export in open formats
Demo coming soon
-
Build custom AI agents
Agent Studio
Build an agent for a workflow that is yours alone: the tools it may use, the data it may read, the approval it must wait for. The guardrails are the platform's, not something you configure from scratch.
- Scoped tools and data
- Approval gates built in
- Every action attributable
Demo coming soon
-
Ready-made agents you deploy
Specialist Agents
Ready-made specialist agents for common operational jobs, run inside Agent Studio under the same approval and audit rules. Deployed, not installed; a capability of the platform rather than a store.
- Common operational jobs
- Same guardrails as custom agents
- No app-store install buttons
Demo coming soon
What you look at
Two surfaces, and a rule about both.
Every figure on this page is invented sample data. We have no customers yet, so nothing here is a result — it is a demonstration of the shape of the thing.
Revenue, next quarter
₹18.4 Cr
+6.2%
High confidence · 14 quarters of history
Demand, next 30 days
−12%
Falling
Medium confidence · seasonality adjusted
Cash runway
47 days
Clear
High confidence · 3 inflows pending
Capacity released
1,240 hrs
68% redeployed
Quarter to date · Human Opportunity Index
Decision trail
- 09:14 Demand signal detected — West region, 3 SKUs Predicted
- 09:14 Procurement reduction drafted for review Recommended
- 09:31 Approved by R. Menon, Head of Supply Chain Approved
- 09:32 Purchase plan updated · vendor notified Executed
Every step is attributable. Nothing executed without the approval above it.
Illustrative example · Mid-market manufacturer — sample data, not customer data.
-
Business systems, as they already are
ERP, CRM, HRMS, the finance ledger, and the spreadsheet three people actually rely on.
- ERP · inventory
- CRM · orders
- Finance · ledger
- Sheets · planning
-
A pattern crosses systems
Repeat orders in one region slow down at the same time as a distributor stretches payment terms. Neither system can see the other. Together they mean something.
-
Demand is likely to fall 12% next month
Stated with its reasoning and how confident the forecast is, not as a certainty.
Medium confidence. Driven by repeat-order slowdown in the West region. Two comparable periods in the last three years.
-
Recommended: reduce the next procurement cycle by 18%
The recommendation carries what it is worth, what it costs if the forecast is wrong, and what it assumed.
Avoids roughly ₹41L of stock that would sit unsold. If demand holds instead, expect a 6-day restock delay on three SKUs.
-
A person decides. Always.
The recommendation goes to the one accountable human with everything they need to disagree with it.
ApproveModifyReject -
The approved decision becomes work
Purchase plan updated, vendor notified, owner assigned, delivery tracked. The approval is recorded against every action that followed it.
-
The business moved before the problem reached the monthly report
Same data. Same people. The difference is that the distance between noticing and acting collapsed from weeks to minutes.
Illustrative example · Mid-market manufacturer — sample data, not customer data.
Explainability
A recommendation you cannot interrogate is just an instruction.
If a system tells an experienced operations manager to cut a purchase order by 18%, the only rational response is "why". If the answer is a confidence score and nothing else, they will ignore it, and they will be right to.
- Recommendation
- Reduce next procurement cycle by 18%
- Confidence
- Medium — two comparable periods in three years
- Primary driver
- Repeat-order slowdown, West region, 3 SKUs
- Assumed
- No new distributor onboarding this quarter
- If wrong
- 6-day restock delay on 3 SKUs; no stockout on the remaining 41
- Data used
- ERP inventory, CRM orders, 14 quarters of history
- Accountable
- Head of Supply Chain — approval required
Illustrative example · Mid-market manufacturer — sample data, not customer data.
Every recommendation carries this. Not on request, not in an export — attached to the thing itself, because a recommendation without its reasoning is not reviewable, and anything not reviewable should not be trusted with a decision.