RECOVERA

Inspiration

Businesses rely heavily on SaaS and cloud services for critical operations. These services are often governed by Service Level Agreements (SLAs) that provide service credits when vendors fail to meet agreed performance commitments.

However, we identified a gap between having a contractual right and actually recovering its financial value. SLA clauses may exist in contracts, while uptime data, incident records, vendor communications, and invoice information are distributed across different systems. A business may be eligible for a service credit but still miss the opportunity because identifying and proving the entitlement requires a manual, time-sensitive investigation.

This problem inspired us to create RECOVERA.

What it does

RECOVERA is an autonomous SLA entitlement recovery agent for SaaS and cloud services.

It connects what the contract promised with what actually happened in the business.

When a potential SLA breach is detected, RECOVERA forms a hypothesis and investigates whether the organisation may be entitled to a service credit. The agent retrieves relevant contractual conditions, analyses service performance, checks incident records, evaluates exclusions such as planned maintenance, searches vendor communications, verifies financial information, and calculates the potential service credit.

RECOVERA then creates an evidence-backed recovery case for human review.

Before any external or financially consequential action, an authorised reviewer can approve, redirect, request additional evidence, reject, or escalate the case.

How we built it

RECOVERA is designed around a state-based agentic workflow.

At the centre of the architecture is an Agent Orchestrator that maintains the recovery goal, current case state, missing evidence, and investigation plan.

A Contract Intelligence Layer uses Retrieval-Augmented Generation (RAG) to retrieve relevant SLA clauses, definitions, service-credit rules, and exclusions.

The agent can dynamically select between multiple evidence sources, including:

A contract repository
A monitoring API for uptime and outage data
A Jira-style incident system
A financial or ERP data source
An approved vendor communication source

An Evidence Graph connects important conclusions to their supporting sources, while a Policy and Confidence Engine controls evidence requirements, uncertainty handling, and action boundaries.

RECOVERA is designed to re-plan when evidence conflicts, a tool fails, a human redirects the investigation, or a vendor provides new evidence.

Challenges we faced

One of our main challenges was ensuring that RECOVERA was a genuine autonomous agent rather than a fixed automation workflow or contract chatbot.

A simple SLA threshold breach does not automatically prove that a service credit is owed. Planned maintenance, contractual definitions, conflicting monitoring results, previous credits, and claim deadlines can change the outcome.

Therefore, we designed RECOVERA to maintain a dynamic case state and choose its next investigation step according to the evidence still required.

Another challenge was balancing autonomy with business control. RECOVERA can autonomously investigate approved systems and prepare a recovery case, but human authority is required before external or financially consequential actions.

Accomplishments that we're proud of

We are proud of designing RECOVERA around a measurable B2B outcome: identifying and recovering contractual value that may otherwise be missed.

The agent does not simply generate an answer. It creates a traceable connection between contractual conditions, operational events, supporting evidence, financial impact, and a proposed recovery action.

We also designed RECOVERA to handle ambiguity, tool failure, human redirection, and vendor denial through re-planning.

What we learned

Through RECOVERA, we learned that building an autonomous agent is fundamentally different from building a chatbot.

An effective agent needs a clear goal, dynamic state, access to tools, evidence-aware reasoning, failure handling, action boundaries, and meaningful human control.

NeuroX 1.0 encouraged us to think beyond AI-generated responses and focus on how autonomous agents can investigate real business problems and produce measurable outcomes.

What's next for RECOVERA

Our current prototype focuses specifically on SaaS and cloud SLA service-credit recovery.

The underlying entitlement-recovery architecture may later be adapted to other B2B contract environments, including telecommunications, logistics, manufacturing, procurement, managed services, and finance operations.

These domains may use different entitlement rules, evidence sources, and calculation logic, but the core RECOVERA model remains the same:

Detect. Investigate. Prove. Recover.