A growing set of procurement vendors are starting to use the word "context." It's worth being precise about what we mean by it, because we mean something different from either of the two patterns the market already has.
Today's procurement-AI vendors approach "context" in one of three ways. Two of them upgrade what's already in your systems. Only one captures the reasoning that never made it into your systems in the first place.
The three patterns
1. Structured-data context
Cleansing, normalizing, enriching, and connecting the data your ERP, P2P, and contract repositories already hold.
- What it does: spend analytics, vendor master normalization, contract metadata extraction, category dashboards, vendor-data enrichment.
- What it doesn't capture: the reasoning behind the decisions that produced the data in the first place.
- Leading vendors: Suplari, Sievo, TealBook, Deducta, Terzo, traditional spend-analytics players.
2. AI assistants and agents over the data
A new generation of tools that put intelligent agents on top of existing systems and conversations: recommending vendors, drafting RFPs, watching vendor meetings, alerting on commitments.
- What it does: predictive pricing, autonomous sourcing journeys, vendor-conversation monitoring, commitment tracking, agent-driven workflow execution.
- What it doesn't capture: a structured record of why a category strategy was chosen, which alternatives were ruled out and on what grounds, who needs to be consulted on which vendors, and the unwritten rules that govern those relationships. These tools generate outputs; they don't persist the reasoning that should drive them.
- Leading vendors: Globality, Arkestro, Abra, and most of the "agentic procurement" entrants.
3. Unstructured-reasoning context — what Cotiss does
Capturing the why behind every decision (from the emails, documents, meetings, and conversations where decisions actually get made) and encoding it into a procurement-specific layer that people and agents can query.
- What it does: decision records, vendor-commitment capture, category-strategy memory, audit-grade reasoning trails, institutional knowledge preserved across people and time.
- What it captures that the others can't: the considered alternatives, the constraints that ruled them out, the stakeholder positions, the vendor history that lives in long-tenured category managers' heads — the substrate that lets any agent (ours, yours, or your orchestration platform's) act with real understanding rather than confident guesswork.
- Leading vendor: Cotiss.
Side-by-side
| Dimension | Structured-data context | AI assistants & agents over the data | Unstructured-reasoning context (Cotiss) |
|---|---|---|---|
| What they upgrade | The data already in your ERP, P2P, and contract systems | The interface and workflow on top of your data | The reasoning that never made it into your systems |
| What they capture | Cleansed, classified, enriched transactional data | Real-time activity signals and agent outputs | The why behind every category decision, vendor choice, and policy exception |
| Where the data comes from | Internal systems of record | Internal systems plus live conversations and agent inference | Systems of record, live conversations and agent inference, emails, documents, meetings, evaluations, contracts, and category-manager judgment |
| What gets lost | Reasoning, history, tacit knowledge | The structured record of why anything was decided | Nothing structural: that is the point |
| How agents consume it | Dashboards and APIs to humans | A vendor-specific agent experience | A queryable context layer any agent or human can call |
| Examples of use | Spend dashboards, contract metadata, vendor master | Autonomous RFPs, predictive pricing, vendor-meeting alerts | Decision records, audit-grade reasoning trails, category-strategy memory |
| Leading vendors | Suplari, Sievo, TealBook, Deducta, Terzo | Globality, Arkestro, Abra | Cotiss |
Why the distinction matters
The first two categories are valuable, but they share a structural limit: neither captures the reasoning that drives procurement in the first place. Cleaner data and smarter agents both assume the reasoning is already somewhere they can read. It isn't. It's in inboxes, in PowerPoint category strategies, in QBR notes, in the head of the category manager who's retiring in eighteen months.
Without that reasoning layer, an agent acting on procurement workflows is fast, confident, and often wrong, routing work efficiently to bad outcomes because nobody told it which vendor relationship is non-negotiable, which clause was hard-won in 2023, or why a multi-source strategy was chosen for this category three years ago.
Cotiss is built specifically for that layer. Not to replace structured data, not to replace agents: to give both something to stand on.