In our experience implementing financial systems for organizations handling sensitive, high volume transactions, spreadsheets usually break at the point where audit pressure, reconciliation errors, and manual controls start creating real risk. Post implementation, the features that matter most are reliable bank connectivity, automated reconciliation, clear approval controls, and reporting that finance teams can trust without manual cleanup. The biggest mistake we see teams make when selecting a TMS is buying for future complexity instead of current reality, underestimating integration effort, and overlooking change management, which often costs more time than the technology itself. Full Name: Arsalan Chauhdary Title: Co Founder and CEO Company: Halo Digital Bio: Arsalan Chauhdary is the Co Founder and CEO of Halo Digital, a software and digital solutions company that helps organizations design, build, and scale secure, high performing digital platforms. With a background in solutions architecture and hands on engineering, he works closely with startups, nonprofits, and growing teams to implement reliable systems that replace manual workflows, improve operational visibility, and support long term scalability under real world constraints. LinkedIn: https://www.linkedin.com/in/arsalan-chauhdary-364944124/ Company Website: https://halodigital.co/
The breaking point I see for moving off spreadsheets is usually a scare, not a strategy: a near-miss on payroll, a debt repayment almost paid from the wrong account, or an auditor showing how one broken macro could hide a cash shortfall. Once leadership feels that risk in dollar terms, the argument for a TMS stops being "efficiency" and starts being "we can't afford this fragility". Post-implementation, the features that keep proving their worth are boring but core: stable multi-bank connectivity so you've got yesterday's cash in one view by early morning, clean ERP integration so reconciliations and postings don't rely on offline files, and a 13-week+ cash forecast with simple scenario switches. Add centralised payments with clear approval rules and a full audit trail and you've covered most of the value; everything else is layering. The biggest selection mistakes I see are buying for edge cases instead of the 3-4 daily jobs the team does, under-scoping integrations because "we'll hook that up later", and treating the TMS as an IT install instead of a change in who does what, when, and with which checks. The software is rarely the problem; weak process design and half-done training are what kill adoption and delay ROI. You can quote me as: "Teams overrate pretty dashboards and underrate plumbing. In practice, cash visibility from bank feeds, clean ERP links, and a forecast you trust matter more than any glossy AI demo. The 'best' TMS in 2025 is the one the treasury team uses every day without heroics, because the data just shows up and the controls quietly work." Josiah Roche, Fractional CMO, Silver Atlas - www.silveratlas.org
Aleksa Baburska, Director of Solution Acceleration at Devox Software. https://www.linkedin.com/in/aleksa-baburska-28501a26b/ Some time ago I led a team to modernize a TMS for a national bank in Europe. Here are some observations as a result of it. What forced the move off spreadsheets The client hit an audit wall: too many manual edits across files were required, and there was no reliable audit trail which was unacceptable for the changes in the supervision board and ownership. To sum up, the decision was dictated by intensive company growth and a far-looking development strategy more than with current operational efficiency. Top 3-5 features that mattered after go-live My experience shows the following benefits of a modern TMS weigh more than others: 1. Stable multi-bank connectivity that delivers predictable and consistent data to be processed (including by AI features). 2. Clean ERP/GL integration for auto-matching and reconciliation, plus payments governance and strong controls (RBAC/SoD, audit logs). The most common mistake was as follows: Teams budgeted for licenses but not real-life handling. So buying the "enterprise suite" before agreeing on the operating model and skipping change management/training is inefficient. What AI in treasury should mean in 2025 As a Director of Solution Acceleration, my primary goal is to implement AI-powered tools into workflows and systems. Practical AI includes explainable forecasting support, huge anomaly detection scope, and proactive exception triage such as missing statements, duplicated payments, unusual cash movements, etc. Before implementing chatbots, a bank needs to focus on the product and its security and reliability. ROI outcomes Once connectivity and reconciliation rules optimized, payment exception handling drops by ~20%. Anomaly detection catches up to 10% alerts more, however this extra workload is minimized by appropriate automated action points and classification.
Context on my experience first: I was a Sr. Finance Manager for FP&A at a $1.6B Business Unit for a major Defense Contractor and worked in Corporate FP&A for 15 years. Now, I run a private investment company and run a stock research firm. I'll answer with a Defense Contractor perspective. "Cash Management for all companies is crucial but for Defense Contractors, there are added complexities, especially if there are large international customers. Negotiated payment terms can be hard to enforce and oftentimes International customers have large milestone payments that can vastly impact the quarter results. Having insight via a centralized, automated dashboard tool is critical. Tools that can feed from the ledger and populate latest transactions helps the business work cash real time."
The biggest mistake teams make when evaluating a TMS is over-indexing on features instead of visibility. What actually mattered post-implementation was reliable bank connectivity, real-time cash positions across entities, and forecasts that updated automatically as transactions changed—not static models that needed constant rework. In 2026, AI in treasury shouldn't mean black-box predictions; it should mean continuously classifying transactions, flagging deviations, and keeping cash visibility accurate enough that teams can make decisions without reconciling spreadsheets first.
With a background in implementing and selecting several TMS and related systems, the major challenges our clients face are lost opportunities due to unmanaged strategic reserves, overhead from manual reconciliation across multiple bank accounts, reduced confidence in cash forecasts, and the inability to scale confidently. The most important features are: multi-bank connectivity; historical cash forecasting data and anomalies; the ability to notify and take corrective actions with human-in-the-loop and governance processes; and the ability to run bank feeds in different formats as required by other systems and channels. The importance of ERP or other ancillary systems integration can vary depending on the organization's complexity and architecture. The most common mistakes when selecting a TMS are buying without considering the enterprise context and failing to align workflows with the change model. Without consensus on the target operating model, the selected TMS systems struggle with adoption, inaccurate forecasts, and expensive subscriptions that fail to deliver business value. I have approached it differently: before jumping to new solutions, the teams need to be aligned on core objectives and critical success factors. Once agreed, then selection follows, aligned to the new target operating model. Unfortunately, without a fully baked model, surprises arise during implementation, integration, and adoption. In my view, bank connectivity is absolutely critical with multiple banks, including virtual ones. This helps diversify risk within a single bank account due to technical and process limitations. AI in treasury should enable autonomous agents to trigger notifications and initiate corrective actions, with a human-in-the-loop. Depending on the organization's maturity, whether they perform forecasting and scenario modeling within TMS or in another software such as FP&A, the needs can vary. But TMS should retain historical data and provide it in formats relevant to the organization. For global orgs, it might be even trickier, depending on organizational alignment and whether they maintain a central treasury function or one per operating entity. The architecture and requirements could vary depending on TMS's role in a given organization. Sam Gupta CEO and Principal Consultant ElevatIQ https://www.linkedin.com/in/samguptausa/ Quote: Selecting a TMS requires realigning your organizational processes, defining their scope, and mapping dependencies.
What finally forces teams off spreadsheets is risk—missed cash visibility, audit pressure, and the inability to forecast accurately once you have multiple entities or bank accounts. Post-implementation, the features that actually matter are reliable multi-bank connectivity, clean ERP integration, real-time cash positioning, and controls around payments and approvals; everything else is secondary. The biggest mistake I see is buying a 'feature-rich' TMS without fully scoping integrations or change management—implementation success is less about software and more about data hygiene and ownership. In 2025, AI in treasury should mean better forecasting, anomaly detection, and alerts—not black-box predictions or marketing fluff.
At Titan Funding, we had to ditch spreadsheets because the transaction volume got out of hand. Manual tracking became a time bomb, with mistakes slipping through, especially during audits. Connecting our bank and ERP systems directly eliminated all those late-night double-checks. My best advice is to focus less on the features and more on training everyone properly. That made a bigger difference than any new tool.
Full name: Ahad Shams Title & company: Founder, HeyOz LinkedIn: Available on request Publish-ready quote (2-4 sentences): "We replaced spreadsheet-based treasury tracking when forecast errors started affecting hiring and marketing decisions. The breaking point was realizing we were spending more time reconciling numbers than making decisions. A TMS only paid off once it gave us daily cash visibility and reduced decision latency, not because it had more features." Response: At HeyOz, spreadsheets failed once transaction volume and variability increased. Multiple bank accounts, subscription inflows, ad spend, and contractor payments meant forecasts went stale almost immediately. The trigger to move off spreadsheets was not audit pressure, but operational risk. We were making six-figure spend decisions with week-old data. Post-implementation, three features mattered far more than the rest. First was reliable bank connectivity with frequent refreshes. Anything less than daily data is effectively historical. Second was clean integration with billing and payroll systems so cash forecasts reflected real behavior, not assumptions. Third was scenario modeling. Being able to quickly see best, base, and downside cash positions changed how leadership discussed growth and runway. The most common mistake I see teams make is buying an enterprise-grade TMS before they have the operational maturity to support it. Over-scoping integrations and under-investing in change management leads to shelfware. A close second is confusing AI marketing with real value. In 2025, AI in treasury should mean anomaly detection, variance explanations, and forward-looking scenarios, not black-box forecasts. Implementation was faster than expected, roughly eight to ten weeks, but only because we limited scope and prioritized data accuracy over customization. If I did it again, I would spend more time upfront aligning stakeholders on how forecasts would be used, not just how they would be built. Optional context: SaaS company, early to mid-stage. Forecast update time reduced from days to minutes, with materially fewer cash surprises.
Because our business is large (with many units and accounts) and risky, we stopped utilizing spreadsheets to track cash. Hand reconciliation was tedious, especially at the end of the month and when making projections. Cash knowledge always lagged. Not because it wasn't working, but because hand-checking was hazardous. Our new plan made connecting to our ERP system and many banks easier. We could monitor real-time cash flow. We made superior predictions since our financial system and model received bank account data instantly. Dashboards were less important than bank network reliability and data refresh frequency. When using technology for finance, teams' worst mistakes are not managing change adequately and choosing too many options. Deviance projections, scenario modeling, and odd things should be part of "AI in Treasury" by 2025. The Treasury shouldn't trust unsubstantiated AI claims.