How AI is changing treasury

Insights within Treasury Operations - interview with Konstantin Khorev, interim Treasury Manager

Artificial Intelligence (AI) is making a significant impact across industries, and corporate treasury is no exception. To explore how AI can be adopted effectively in treasury, we spoke with Konstantin Khorev, a finance and technology expert with extensive experience in the field. Drawing on his time at a leading global environmental commodity trader and climate solutions provider and broader roles in this sector, Konstantin offers an insider’s view into what it really takes to bring AI into treasury processes. With a PhD in mathematics and a strong background in technology, data analytics, and system integration, Konstantin is at the forefront of AI-driven treasury innovation.

The business case

A prominent global player in environmental commodity trading and climate solutions, has integrated AI into its treasury operations not merely for the sake of innovation, but due to its tangible benefits.

From idea to implementation: A complete journey

The adoption of AI in treasury begins not with technology, but with a clear business need. The Treasury team identified forecasting as the area with the most potential for AI-driven improvement, particularly for medium- to long-term cash flow and FX exposure forecasts. Konstantin explains: “Reliable forecasting supports better liquidity planning and risk management. Whether you’re managing large working capital cycles or volatile currency exposures, better visibility means better control.”

The AI journey followed a structured, end-to-end process:

1. Business case detection – Forecasting cash flow beyond the short term is critical in capital-intensive industries like commodity trading. While short-term cash positions are typically predictable based on known payables and receivables, medium- and long-term visibility presents a challenge. AI is well suited to tackle this.
2. Data gathering and cleaning – Quality data is Konstantin emphasizes that no model, however advanced, can compensate for poor input.
3. Initial data analysis and model search – The team tested both parametric and non-parametric models to identify the best fit.
4. Back-testing – Only after robust testing were the models prepared for One model, FX exposure forecasting, is fully operational, while the other, accounts receivable forecasting, is still in back-testing.
5. Deployment – The models were designed for simplicity. They run locally on a standard laptop, outputting results into Excel for easy use across BI dashboards and treasury systems. There was no business need to embed them deeper into the IT

A key insight is that no specialized hardware, complex software, or large team was required.

This was a one-person effort, feasible because of strong technical skills and powerful open-source libraries. However, the true challenge lay elsewhere.

“Cross-functional alignment is just as important as technical implementation”, Konstantin stresses. Treasury had to communicate the model’s dependency on upstream data quality and work closely with senior management, IT, and business teams to ensure adoption and long-term success.

Lessons from implementing AI: A treasurer’s guide

Konstantin shares several learnings from this and other AI implementations in treasury:

• Start with a real business problem. AI for AI’s sake rarely delivers value. Our objective was improving accuracy in cash forecasting and FX risk assessment, not showcasing technology.
• Data quality is non-negotiable. If your input data comes from systems not controlled by treasury or is poorly structured, model reliability suffers.
• Stakeholder trust is essential. AI introduces a ‘black box’ effect that may raise eyebrows. Explaining the model logic and involving users from the start helped build
• No need for big Modern AI models can be built with Python, open-source libraries, and a laptop. The bottleneck is expertise, not hardware.
• Iterate We didn’t wait for a perfect environment. The models improved through use, feedback, and refinement.

Tangible business value

The implementation of AI improved multiple KPIs. For cash forecasting:

• Idle cash levels dropped, freeing up capital for trading
• Missed trade opportunities due to liquidity uncertainty were reduced
• Management satisfaction improved, though harder to quantify

In FX risk, hedge effectiveness became more measurable and consistent.

For any commodity trading company, every cent counts. Optimizing cash flow and risk exposure through AI is not just a technical gain. ‘’It is a competitive advantage,’’ Konstantin says.

Other opportunities for AI in treasury

Beyond cash flow and FX exposure forecasting, AI can also enhance other key treasury functions:

• Fraud detection and risk management. Machine learning algorithms can detect anomalies in payments and transactions, identifying potential fraud in real time.
• Automated reconciliation. AI solutions can match transactions across multiple accounts and systems, drastically reducing reconciliation time.
• Dynamic liquidity management. AI can optimize liquidity allocation based on real-time cash positions, enhancing yield and minimizing idle cash.

“These applications show that AI is not just a theoretical concept. It is actively improving treasury operations,” Konstantin adds.

Looking ahead: AI in treasury by 2030

Konstantin envisions a future where AI becomes fully embedded in treasury workflows. “By 2030, we will see AI-powered autonomous cash management solutions and AI-driven decision-making. Treasury management systems will incorporate natural language processing (NLP), enabling treasurers to interact with them conversationally. Instead of manually inputting data, treasurers might simply instruct their TMS to hedge their FX exposure using specific instruments, and the system will execute the task.”

The integration of generative AI will be another game-changer. “While treasury is not the primary focus for AI developers today, we can expect AI to automate complex valuations, trade finance and other tasks. AI-driven assistants will significantly reduce manual workload, allowing treasurers to focus on strategic initiatives.”

Regulatory compliance will also evolve alongside AI. “With increased AI adoption, regulators will demand greater transparency. Treasurers will need to ensure that AI models are explainable and comply with evolving regulatory standards.”

Final thoughts: Embracing AI in treasury

For treasurers considering AI adoption, Konstantin offers practical advice. “Start small but think big. Focus on quick wins to demonstrate value, but always have a long-term vision for how AI can transform your treasury function.”

He emphasises to treasurers that AI is an enabler, not a replacement. “AI is not here to replace treasurers. It is here to empower them. Those who embrace AI will gain a strategic edge in an increasingly data-driven world.”

AI is no longer a futuristic concept. It is shaping the future of treasury today. As more treasury teams adopt AI, the role of technology in corporate finance will continue to expand and evolve.