Symphony is an agentic operating system: AI agents that understand your business, work across your applications and carry out tasks under your control. Its first edition, Orchestra, puts specialised agents to work on procurement, connected to procure.az through its MCP server.
Finishing one task means carrying information between documents, conversations, databases and applications — by hand, every time.
A chatbot can tell you what to do. Someone still has to open the systems and do it.
Handing work to AI usually means trusting it with everything or nothing, with no way to hold back the actions that need a person.
A tool tied to a single model makes every task pay the same price in quality, speed and cost, whether it needs to or not.
Create agents for specific roles and processes. Define what each one should achieve, which tools it may use and when it must ask a person.
Connect policies, documents, product information and operational data, and give each agent only what its task needs, within your permission rules.
Agents retrieve records, prepare documents, update information and trigger workflows through the integrations you approve.
Bring agents, business rules and human approvals into one process, with defined handoffs, dependencies and escalation paths.
Choose the model per task to balance quality, speed and cost, while every workflow keeps the same business context.
Every agent action is logged. Review activity, inspect results, evaluate performance and require approval for sensitive actions.
The procurement edition: specialised agents for RFIs, RFX processes and procurement analytics, working with procure.az data and actions through its MCP server, with human review at defined decision points.
| Assist | Approve | Execute | |
|---|---|---|---|
| What the agent does | Retrieves information and prepares a recommendation | Prepares an action for a person to review | Completes authorised tasks within the limits you set |
| Who decides | A person, using the agent’s recommendation | A person approves before anything runs | The agent, inside its limits — exceptions go to a person |
| Good for | New workflows and sensitive information | Actions with financial or legal effect | Routine tasks whose results are already proven |
With Orchestra, prepare RFIs, structure supplier questions, review submissions and analyse procurement data on procure.az, with a person at every decision point.
Check account details, resolve routine requests and hand complex cases to a person with the full context attached.
Prepare meeting briefs and proposals from your own records, check invoice information and draft follow-ups for review.
Monitor exceptions, answer internal questions, route service requests and prepare status reports.
A chatbot talks. Symphony agents also act: they plan the steps, use your tools and track a task through to completion.
Each workflow runs at the level you choose — assist, approve or execute — and you widen it only as results are proven.
Orchestra works with procure.az data and actions through its MCP server, so procurement agents start on the platform your teams already use.
Review every agent action in the activity log and evaluate each workflow on completion, accuracy and how often people stepped in.
From the Greek symphōnía, “sounding together”. In a symphony every instrument plays its own part, and the result is one piece of music. Symphony does the same with agents: each has its own role and tools, and together they finish one piece of work. Its procurement edition takes the name of the players themselves — Orchestra.
Symphony agents act only through the tools and permissions you grant. Sensitive actions can wait for a person’s approval, every action is logged, and you decide — workflow by workflow — how much autonomy the agents get.
Allmaz is the AI product studio of Smart Solutions, which built and operates Azerbaijan's unified public procurement portal — established by presidential decree and delivered as one of the country's first public-private partnerships in digital government.
Bring us one workflow worth improving. We will map the process, show where agents can help, and agree what a successful pilot must demonstrate.