See what buy-side reconciliation teams are saying about the future of their work
Fragmented Notice Formats. Disconnected Data Sources. Break Queues That Never Quite Shrink.
These are familiar operating conditions for most buy-side reconciliation teams, whether they use automated systems or not. As instrument complexity and data inconsistency continue to increase, rule-based matching engines are generating more exceptions than they can resolve. The rules are still correct. But on its own, a rule library was never going to be able to keep up with every new notice format and instrument type.
The significant gap between “automated” and “actually keeping up” came through loud and clear in a live poll we ran during a recent webinar about deploying agentic AI for reconciliation. The results described a process that is certainly more efficient than manual spreadsheets but still falls short of where it needs to be.
Catching Up To Complexity
Most of our poll respondents are still primarily using manual or spreadsheet-based processes for reconciliation, with the rest split between using a rules-based matching engine by itself and using a rules engine with automation layered on top. None have AI agents in production yet, which aligns with the broader market. The real challenge, however, is that fragmented notice formats and disconnected data sources are generating new types of exceptions faster than teams can write rules for them.
Ask a rules engine to match anything it hasn’t seen before, and it kicks the item into a break queue instead of resolving it. Now throw in complex instruments, privately held positions, and multiple counterparties sending notices in different formats. The resulting queue keeps growing even when the underlying process is technically “automated.”
Where Does The Time Actually Go?
The second question we asked was about what slows recon down the most. The answers split three ways:
- The sheer volume of breaks to investigate
- The ongoing work of configuring and maintaining matching rules
• Quarter-end volume spikes
All three of these answers point to the same root cause: it’s more of an unstructured and fragmented data feed. Data extraction is a bigger problem, which we are trying to resolve with an LLM. A rules engine can only handle what it’s been explicitly told to expect. Every new break pattern means the recon team has to write a corresponding rule, test it, and deploy it in a continuous cycle of maintenance that takes time away from clearing the current queue.
This is exactly the gap that agentic AI is designed to close.
For example, the IVP Reconciliation Solution pairs a rules-based matching engine with an agentic intelligence layer that reads unstructured notices, proposes matches that rules don’t yet cover, and routes actual exceptions with a recommended resolution.
The Agent Notice Processing AI agent ingests fragmented notice formats into a common structure before running a single rule. From there, the Intelligent Suggestions agent recommends the next step for investigating breaks, extrapolating from how similar breaks were resolved previously. As a result, the analyst always starts with a proposed action instead of doing the grunt work of identifying the potential matches and forcing them to match
The Trust Question, Answered By The Room
The most direct signal came from the third poll question: How comfortable is your team letting an AI agent auto-approve a clean reconciliation?
Most respondents landed on the same answer: yes, as long as there are thresholds and an audit trail. The rest said they would only go as far as shadow mode, where the agent can make suggestions but never acts alone. Interestingly, nobody chose “humans approve everything” or “ready for full autonomy,” the two extremes of this scale.
None of these answers is wrong. The split itself is the point. Ops and compliance teams understand that AI agents can do the work, but they need reassurance that the system can show its reasoning when a regulator, auditor, or portfolio manager asks why a specific break was cleared. That’s more of a governance question than a capability question, and it’s the one a hybrid model is designed to answer.
In the IVP Reconciliation Solution, for example, human oversight isn’t bolted on after the fact. Every match and auto-clear carries a full audit trail back through the decision, which is what lets a firm achieve full autonomy for clean, repetitive matches and human sign-off whenever the agent’s confidence or the item’s risk profile calls for it.
The approval step runs through a dedicated sign-off workflow, so a firm can dial the model from shadow mode to bounded autonomy without changing the underlying rules or losing the audit trail of who approved what.
Closing The Gap
Line up the three sets of poll answers, and a clear picture emerges. Teams running rules-based automation are buried in breaks and tired of maintaining rule libraries by hand, but they prefer a human checkpoint before any break is cleared.
That’s exactly the gap between a rules-only engine and an agentic approach. Rules stay accurate and auditable for the types of breaks they’re designed to catch, while AI agents pick up the notice formats, break patterns, and repetitive resolutions that rule engines simply can’t cover. All without asking a firm to add headcount or write another six months of rules.
What This Means Going Forward
For firms still operating in manual or semi-automated reconciliation setups, rising notice volumes and increasing instrument complexity will continue to test the limits of traditional workflows.
Keep in mind, however, that while a shift to agentic AI will help clear breaks faster, the real emphasis is on giving the reconciliation team access to the audit trail and explainability that regulators, auditors, and internal risk teams now expect as the industry standard.
For a more detailed discussion of these issues, watch our webinar, “Beyond the Rulebook: Deploying Autonomous AI Agents for Reconciliation,” to explore how notice processing, agent-driven matching, and sign-off work together in practice.
You can also learn more about the IVP Reconciliation Solution or contact us to talk about what these capabilities could mean for your team.
