How Payhawk Achieves 60% Faster Workflows with AI Agents

Payhawk, the Bulgarian spend management platform that processes corporate payments and expense claims, is launching four AI agents designed to automate finance tasks as part of its strategy to expand AI capabilities for chief financial officers (CFOs).

The agents are part of a shift towards agentic AI, where autonomous software systems can complete tasks independently rather than simply responding to queries. 

Unlike chatbots that provide information, these agents can execute actions such as booking travel, processing payments and managing procurement workflows without human intervention.

Inside Payhawk’s AI plan

Konstantin Dzhengozov, Co-founder and CFO at Payhawk, positions AI as complementary to existing financial controls rather than replacement technology: “AI is a layer on top of all the controls and all the frameworks that already exist,” he says.

The company has developed its AI strategy over seven years, conducting trials that demonstrated measurable improvements in financial operations. 

Early testing showed companies could accelerate month-end closing processes, with Konstantin noting that “companies can close their month much faster, usually two, three or four days quicker than before.”

Each agent operates within existing permission structures and policy frameworks, maintaining audit trails for regulatory compliance. 

All actions performed by the agents are logged to provide accountability and visibility for finance teams, addressing concerns about autonomous system oversight in regulated financial environments.

How Payhawk agents target workflow bottlenecks

The Travel Agent uses natural language processing (NLP), the AI technology that enables computers to understand human speech, to book corporate travel based on employee preferences and company policies. 

The system automatically groups related expenses and generates trip reports that can be approved with single-click functionality.

Payhawk’s Payments Agent addresses support queries by providing insights into failed transactions, blocked payment cards and funding issues. 

The company claims this agent deflects 40% of helpdesk enquiries while suggesting compliant next steps that adhere to relevant regulatory requirements.

The Financial Controller Agent focuses on expense submission workflows, automatically capturing receipts and uploading documents from vendor portals. 

This system flags anomalies in spending patterns and escalates reminders around month-end closing procedures, with Payhawk claiming expenses can be submitted twice as fast.

The Procurement Agent targets purchase request workflows, where the company reports a 60% improvement in procurement time. 

Users communicate requirements through a chat interface, while the agent applies budget constraints and policy requirements before routing approval requests to appropriate personnel.

How enterprise focus drives agent development strategy

Hristo Borisov, CEO and Co-founder at Payhawk, says: “Enterprises don’t need more chat, they need outcomes.”

This focus mirrors broader industry challenges in scaling AI systems beyond experimental implementations. 

Many AI agents lack the security, permission management and audit capabilities required for corporate financial operations, particularly where transaction compliance is mandatory.

Hristo identifies specific enterprise requirements that differentiate Payhawk’s approach from consumer-focused applications: “The majority of agents on the market today lack enterprise capabilities to be adopted at scale, such as permissions, policies, multi-tenancy, audit trails and data security standards – all absolutely critical when it comes to business payments,” he says.

The agent deployment maintains existing organisational hierarchies and approval workflows while automating routine tasks. 

This approach addresses concerns that AI implementation might disrupt established financial controls or create compliance risks in corporate payment processing.

Multi-tenancy capabilities allow the system to serve multiple clients while maintaining data isolation, a requirement for software providers handling sensitive financial information. 

The audit trail functionality provides detailed records of agent decisions, supporting regulatory reporting and internal control frameworks.

“Our AI agents act within your controls and finish real finance tasks, so the easy thing for employees is also the right thing for the business,” Hristo concludes.

AI across industries 

To learn more about the advantages, disadvantages, challenges and hurdles of AI in fintech, leaders in the industry will convene for a panel titled ‘AI in Fintech’ at FinTech Live London. Tickets are available here. 

Agentic AI will also be discussed by procurement software giant GEP  at Procurement & Supply Chain Live London. Register your interest here.