Articles
Build vs. Buy: Why Treasury Teams Are Taking Custom TMS Development Seriously Again
- By AFP Staff
- Published: 11/17/2026

For decades, building a custom treasury management system (TMS) has been the impractical option due to cost, limited staff and a shortage of technical skills. However, advances in technology, including AI-assisted development, have reopened the build-versus-buy decision.
“We haven’t seen this happening in 25+ years,” said Tom Hunt, Director of Treasury Practice, AFP. “With AI, it’s now possible.”
Faced with technology gaps and unmet promises, some leading treasury teams are building their own in-house solutions designed around their specific needs.
In a recent Association for Financial Professionals discussion, treasury executives described how they are putting that option into practice, from building a custom TMS to developing tools for bank connectivity and account management. Along with other examples from treasury teams, their experiences illustrate both the possibilities and the commitment involved in taking ownership of treasury technology.
Key Takeaways
- Build versus buy is back on the table. Custom development is becoming a viable option for teams whose needs are not being met by commercial systems.
- Building can take different forms. Teams are developing complete systems, creating targeted tools and combining third-party connectivity with custom workflows.
- Greater control requires an ongoing commitment. Teams need to weigh the benefits of customization against the work of designing, implementing and maintaining a custom solution.
- Specific requirements and reliable data should guide the decision. Evaluating real use cases can expose technology gaps, while dependable connectivity and consistent data definitions support either approach.
The work involved may be the deciding factor right out of the gate
Being able to design a TMS around your treasury team’s requirements is the appeal of a custom build. But whether that flexibility justifies the work involved in building and maintaining it depends, in part, on how well an existing system can meet those needs.
A 2024 AFP article about The Heritage Group’s TMS search illustrates the tradeoff. The team considered a custom solution from a software consultancy that would meet its specifications and come in under budget. But treasury would have had to define everything it needed, and future enhancements would cost extra. For these reasons, the team chose an existing platform that met its must-haves and included routine maintenance and upgrades.
For other treasury teams, the gaps in available systems are making the development commitment worthwhile. One executive in the AFP discussion described approaching multiple vendors about a custom TMS: “Each one said it could be done. None delivered.”
Building a custom TMS
One executive in the AFP discussion said his treasury team now has a custom TMS, built in-house using a cloud computing platform, with a cash sub-ledger for each of the company’s entities. Cash pooling and intercompany payments are up and running, and all intercompany activity is handled digitally rather than through transfers of funds.
The team is building out daily automatic cash reconciliation and bank interfaces, where the homegrown solution will retrieve bank data and bring it into the system. While they’re still using the existing TMS vendor for intraday bank information and some payments, the plan is to bring those functions in-house as well. Additionally, banking partners are required to exchange data through the custom TMS.
After five years of development, the system is approximately 70% complete. While the project has taken longer than expected, the team expects the additional time to be worthwhile.
Combining third-party infrastructure with custom tools
Building doesn’t have to mean developing every component internally. Bandwidth, a finalist for the AFP 2026 Pinnacle Awards, combined third-party bank connectivity with tools designed by its treasury team to replace a legacy TMS.
Bandwidth’s previous system only connected to the company’s primary banks, leaving treasury to manually consolidate data and perform most of the reconciliation. Adding functionality meant purchasing and implementing additional modules, which was a problem for a lean team.
Instead of purchasing another TMS, Bandwidth decided to use Atlar’s Claude Model Context Protocol (MCP) connector as the AI operating environment. They then used Claude to develop reporting tools and workflows around the team’s requirements.
The resulting platform provides a consolidated view of cash across entities and banking partners, along with automated forecasting and liquidity analysis. The team also developed tools for multilateral netting, debt monitoring, covenant tracking and bank fee analysis.
The shift was as much about mindset as technology. Rather than approaching the project as a search for treasury software, the team approached it as an infrastructure problem it could solve by combining third-party connectivity with tools built around its own requirements.
Building for bank connectivity
Custom development can also be narrower in scope, addressing individual processes without the need to replace an entire TMS. Executives in the AFP discussion talked about their efforts to improve bank connectivity and reduce the work involved in managing bank accounts.
One treasury executive found three or four separate connections to the company’s banking partners — each developed independently. In one instance, a bank revealed that it was working with a business partner on an arrangement treasury knew nothing about.
The company’s entrepreneurial culture had encouraged people to act independently. But as concerns about AI and cyber breaches grew, the company began moving toward centralized ownership. Its internal tech partner is now building a single interface to collect the banking data that serves the whole organization.
Building to handle KYC issues
Opening a bank account used to take three days. Now it takes three to six weeks, according to the group’s experience. “KYC is the biggest problem; ask any treasurer,” said one treasury executive.
Much of the information banks require already exists in incorporation records, government filings and the banks' own payment histories. Yet companies are often asked to provide the same information each time they open a new account. For organizations with thousands of accounts, managing signer information, documentation requirements and FBAR filings can cost millions of dollars annually in outside support.
To address this, one company built an AI agent-based product, in partnership with one of its banks, that opens, closes and amends bank accounts. By automating much of the manual work, the company is seeing promising results in addressing the KYC burden while freeing its treasury team to focus on higher-priority tasks.
Ask yourself these three things before you build or buy
The experiences discussed here suggest that building custom treasury technology warrants consideration. They also show why the decision needs to extend beyond the feasibility of the tool’s creation.
Here are three things to ask yourself if you’re considering whether to build or buy:
- Name the specific requirements of the system. The executives who contributed to this discussion said they learned the most when they moved beyond standard demos and asked, “How will a solution handle our use cases?” The discussions that followed exposed gaps that weren’t apparent in the initial presentation.
- Identify the ongoing time commitment. It’s very important to consider how much time the design, implementation and maintenance will cost your team alongside functionality. Greater control comes with ongoing responsibility for the solution.
- Establish reliable data and connectivity. Whether you decide to buy a platform or develop your own, consistent definitions and dependable data are essential. AI can accelerate the workflow, but fragmented processes and poor-quality data will undermine the results.
Building is becoming a practical option for some treasury teams, whether that means a complete TMS or a tool for a particular process. For anyone considering the build-versus-buy decision, the first step is to determine whether custom development offers enough value to justify owning the work, or whether an existing solution already meets your team’s needs.
Copyright © 2026 Association for Financial Professionals, Inc.
All rights reserved.

