Charles Bennett (TIS): foundations before AI
Charles Bennett, Chief Product Officer at TIS, separates real value from marketing in treasury AI and explains why connectivity remains the true foundation. Interview by François Masquelier (ATEL).
Where does AI actually add value in treasury today, and where is it still just marketing?
The most tangible example is cash forecasting. Machine learning models can analyze historical payment behavior, seasonality and external drivers to generate accurate forecasts that improve over time. AI also delivers clear value in data aggregation, anomaly detection, scenario modelling and stress testing at scale, while improving operational efficiency by reducing manual, repetitive tasks. These are structured processes where AI can continuously learn and improve. Where AI is still often marketing is where it sits disconnected from the underlying data foundations. When forecasting solutions are applied to fragmented or low-quality datasets, AI does not fail because of the model itself, but because the inputs are inconsistent, incomplete or poorly connected. The result is often unreliable outputs, which proves worse than no automation at all. The gap we see across the market is not a lack of AI capability, but a gap between AI enthusiasm and AI readiness, driven by data quality, skills and governance challenges.
What new risks does AI introduce in payments and cash management?
AI does not just introduce new capabilities; it introduces new layers of risk that treasury must actively manage. The first concerns data and model integrity: AI systems are only as reliable as the data they are trained on, and fragmented or incomplete treasury data can produce outputs that appear precise but remain fundamentally flawed. The second concerns explainability and audit. Many AI models function as black boxes, which creates challenges around auditability, regulatory compliance and internal governance. My teams must be able to explain why decisions were made, not just what the output was; without that,
adoption stalls through lack of trust, or tips into over-reliance. The third concerns data security and vendor risk. As AI platforms require access to sensitive financial and payment data, they raise data isolation and vendor accountability concerns. Treasury must ensure that AI adoption aligns with strict data governance and security frameworks rather than working around them.
"AI does not solve these problems; rather it amplifies them if they already exist."
Which treasurer tasks will disappear within five years, and which can never be automated?
AI will not replace treasury; it will rebalance how time is spent. Manual and repetitive tasks such as data collection from banks, ERPs and TMS, data consolidation and formatting, and routine variance checks are likely to decline significantly. These are precisely the tasks where AI can operate faster, at scale and with greater consistency. The core of treasury, however, will not be automated, because it depends on judgment, context and accountability. Strategic liquidity decisions, interpretation of anomalies in business context, governance, compliance and audit accountability will remain fundamentally human. AI can inform decisions, but it cannot own them. Strong bank connectivity stays the most important foundation, because modern treasury depends entirely on data completeness, quality and timeliness. The organizations that succeed with AI are not those that adopt it fastest, but those that deploy it most deliberately, starting with the right foundations.
Why a strong and robust bank connectivity remains the most important foundation for an excellent treasury management?
AI, and modern treasury more broadly, depends entirely on data completeness, quality, and timeliness. Treasury teams that invest in connectivity first are able to move faster from AI pilots to production, generate reliable insights faster, and build trust in automated outputs. As we consistently see, the differentiator is not the sophistication of the AI—it is the strength of the foundations underneath it: data, connectivity, and governance. AI does not solve these problems; rather it amplifies them if they already exist. The organizations that will succeed with AI in treasury are not those that adopt it fastest, but those that deploy it most deliberately, starting with the right foundations.