NeuGroup Quick Takes: Capital Markets, Global Cash and Banking

# Technology

# AI

# People and Talent

Editor's note: NeuGroup Quick Takes will be updated with new takeaways following the completion of each of NeuGroup's fall peer group meetings.

Event Week: October 6-8, 2026, Atlanta NeuGroup for Capital Markets (Oct. 6-7), Atlanta. Sponsored by Burnham Sterling NeuGroup for Global Cash and Banking (Oct. 7-8), Atlanta. Sponsored by Ripple Assets under construction don’t have to sit on the balance sheet. Large investments such as manufacturing plants, real estate and aircraft can spend years on the balance sheet before they produce revenue, dragging down return metrics even when the investment itself is sound. At the NeuGroup for Capital Markets meeting, Burnham Sterling principal Erik Lange posed questions to a finance leader from a global logistics company who described avoiding that balance sheet drag by using a capex financing structure arranged by Burnham Sterling. A third party holds the asset under construction on its books while the company has the asset constructed to its specifications.

The company pays interest to the third party on the funds advanced during the asset’s construction or development, at rates comparable to its own bond cost, plus legal and setup fees for the structure.

The asset isn’t recognized on the balance sheet until construction or development is complete, when it is recorded as a finance lease asset with a matching liability that is reflected as a non-cash financing activity rather than capex. The company takes title at lease maturity (the end of the financing period).

The structure works only if the corporate doesn’t own or control the asset before the closing of the financing, so the company vetted the structure with its internal accountants and received sign-off from two successive audit firms.

Automation keeps a crisis-era cash forecast from going stale. When a labor disruption sharply cut revenue at one global manufacturer, the CFO asked how long its liquidity would last. The corporate controller, whose job also covers treasury, built a 13-week cash forecast in a spreadsheet that took four to five hours a week to roll forward. That was too much to sustain for the resource-restricted finance team: Once the crisis passed, the team stopped updating it and rebuilt it only when leadership asked.

At the NeuGroup for Global Cash and Banking meeting, in a session led by Teala Clay , Ripple Treasury Vice President, Global Financial Risk & Market Strategy, two members of the company’s controllership team described turning that stopgap into a standing process by implementing Ripple Treasury , formerly known as GTreasury.

The rollout began this year and already connects more than 20 banks across 10 countries. The system now books journal entries for bank transactions automatically, with any new or unusual payments reviewed and tagged so they post correctly the next time.

About five months in, the weekly forecast now takes under five minutes, and the team has cut a day from the month-end close for seven entities. The team uses the time saved to test forecast accuracy and expand to a longer time horizon, and they expect savings to grow as the solution goes live in additional regions.

Staying flexible in an unpredictable rate environment. Capital markets members arrived as expectations swing from rate cuts to a prolonged hold or hikes under Fed Chair Kevin Warsh. Some said they doubt high rates can stick in a midterm election year, while others pointed to hyperscaler borrowing pushing long-end yields higher.

One member said that uncertainty makes higher-for-longer less a rate call than “a risk management and optionality” problem. She looks first at internal levers that work in either scenario:

One is working capital optimization, such as using supply chain finance to extend payment terms to suppliers. Cash the company holds onto longer is cash it doesn’t have to raise in the bond or loan markets, which matters most when borrowing is expensive.

Another is scaling back buybacks. With rates high, cash spent repurchasing shares gives up more interest it could earn or borrowing it could avoid, and with stock prices already at high multiples, she said the return on buybacks becomes less compelling.

Another member’s company has shifted to sizing its floating-rate debt against its cash portfolio, after swaps put on in 2020 and 2021 purely for carry later performed poorly.

That same company averages into rate locks over the quarters before each planned issuance, so it never bets on a single rate view. But another member advised that this approach works only if leadership and the board agree upfront that locks are hedges; otherwise, losses invite explanations and can become a career risk.

An AI agent’s first big mistake may be its last. One capital markets member’s FX and commodities team has an agent that tells traders each morning which exposures, tenors and volumes to hedge, checks counterparty limits and drafts the messages to banks. It has worked well enough for the team to move off its spreadsheets, but a trader still places every trade. The member said he worries that one bad trade, such as hedging far more copper than the company needs, could cause them to decommission the agent, as he doubts leadership would give it a second chance.

Most of his peers haven’t settled who would answer for that kind of error: 41% said in a live poll that accountability for an agent’s mistake is unclear at their company.

At the global cash meeting, NeuGroup’s Ron Chakravarti noted that treasury has handed off work before, to shared service centers, managed service providers and banks, but each came with a contract, a service-level agreement and someone accountable.

Agents arrive with none of that, while a bank treats any instruction released with a company’s credentials as the company’s own. His advice was to decide for each process whether an agent may prepare the work, release it or never touch it, and to settle who owns the outcome before the agent goes live.

Four NeuGroup meetings brought members together across the country for discussions spanning regional treasury, cash investing, leadership and AI in finance. Members shared what they’re trying, what’s working and where they’re running into challenges—from getting regional treasury a stronger voice at headquarters to figuring out where AI can help without sacrificing judgment and expertise:

NeuGroup for US Regional Treasury & Finance Leaders (Sept. 28-29), Atlanta. Sponsored by EY and Deutsche Bank

Women in NeuGroup (Sept. 29), San Francisco. Sponsored by CNote and R. Seelaus & Co.

NeuGroup for Cash Investments (Sept. 30-Oct. 1), Pleasanton, CA. Sponsored by Invesco

AI in Finance Transformation Pop-Up (Oct. 1), San Francisco. Hosted by Anthropic

Regional teams want a seat at the table. Getting one is another matter. For regional treasury leaders of non-U.S.-based companies, being viewed as strategic business partners rather than primarily operational teams can determine how early they are brought into decisions made at headquarters. A poll at the meeting showed how inconsistent that involvement remains: 88% of respondents said whether HQ consults the U.S. team before making a decision depends on the issue, while none said proactive consultation is the typical model.

Members discussed how regional leaders can build influence by bringing local expertise to HQ, pushing back when needed and flagging issues before they become problems. Personal relationships matter too: direct access to decision-makers and regular communication can help prevent the kind of information gaps members described, including instances when U.S. teams were surprised by developments HQ assumed they already knew.

Make the banks help connect the dots. One assistant treasurer suggested a practical way to improve visibility between headquarters and regional teams. He requires six of the company's top-tier banks to send him a monthly email summarizing whom they spoke with across the company and what they discussed. The reports give HQ a view into conversations happening in the regions and help identify when it may need to get involved.

The practice also puts some responsibility on banks to coordinate coverage across regions—an area where companies often find their banking partners fragmented.

Don't automate away the work that teaches people to think. As AI takes over more routine work, members of Women in NeuGroup (WiNG) raised a longer-term talent question: How will junior employees develop the judgment that more experienced professionals built by doing that work themselves? One approach discussed was giving employees AI-generated work and asking them to find where it is right, where it is wrong and what they would change, turning the technology into a tool for developing judgment rather than bypassing it.

The limits are already showing up in live use cases. One WiNG member said a team of AI agents built for the 13-week cash forecast is about 80% of the way there, but people still need to validate the underlying data before the forecast can be published or reported to leadership. That remaining 20% is where knowing the process well enough to spot a problem still matters.

A similar point surfaced at the Anthropic event. One member suggested separating tasks where the finished product is what matters from those where creating it is part of the learning. AI might produce a strong design document, for example, but writing one can be how an employee develops a deep understanding of the underlying system. The lesson: automate aggressively, but recognize when the process itself is building expertise.

Access to AI isn't the same as having the people to put it to work. Members are increasingly able to experiment with powerful AI tools themselves. Turning those experiments into reliable workflows is a different job. Discussions at WiNG highlighted the advantage enjoyed by finance teams with dedicated systems developers and people who can bridge finance and technology.

One member at the AI transformation pop-up offered an unusually advanced version of that model. The treasury team works with engineering resources as it moves prototypes into production. The treasurer at another company acknowledged advantages most finance organizations don't have: access to leading models, a newly built treasury function and little legacy technology or process inefficiency. The experience underscored that the scarce resource may not be access to an LLM, but access to people who understand both the finance problem and what it takes to build a production-ready solution.

Using AI to help AI help humans. A member of the cash investments group who described himself as nontechnical turned a highly manual interest-income and cash forecasting process into a repeatable generative AI workflow. It brings together investment holdings, CDs and time deposits, expected portfolio flows and treasury-specific assumptions to support an 18-month forecast. Key to his success: He asked a second Gemini instance to help instruct the Gemini workflow he was creating—using AI to build an AI solution. Gemini also helped generate questions around assumptions, safeguards and reinvestment logic. The member then tested the output alongside the existing process for several months before putting it into regular use. The example showed how treasury practitioners who deeply understand a process can use generative AI to automate highly specific work without first becoming developers or AI specialists.

More ambitious AI still requires the right plumbing. Another member described the work required to connect treasury data across multiple systems, give that data the right context and eventually use it for dashboards, counterparty monitoring, forecasting and agent-driven workflows. Note the contrast between the two examples: An individual practitioner can start small and create value with tools available today, while scaling AI across treasury requires much more deliberate investment in data, infrastructure and governance.

Bring the control functions in before the agent goes live. Members looking to automate treasury work are discovering that some of the hardest problems start when an AI agent needs access to bank portals, ERP data and other controlled systems. At one company, discussion of retrieving real-time bank data quickly ran into multifactor authentication and security restrictions; APIs and other controlled connections offered a safer route than finding ways around those protections.

The same issue surfaced at the regional treasury meeting, where members discussed AI applications constrained by access to company data and the need to keep activities such as payment release, accounting entries and audit evidence inside appropriately controlled environments. The takeaway is to involve IT, internal audit and other control partners while designing the workflow—not after a promising prototype is already built.

AI's impact on talent and team construction. One WiNG member recalled initially expecting AI to replace 30%-40% of the team's work, leading the organization to leave vacancies unfilled as employees departed. Instead, the smaller team became stretched and the organization reconsidered its approach: Rather than shrinking treasury teams, the member said AI is more likely to change the skills needed, shifting toward more strategic, value-adding work.

That prompted a broader debate over whether to develop existing employees or hire people with data, engineering or AI expertise. Treasury knowledge helps employees recognize when an output doesn't make sense; technical skills can expand what teams are capable of building.

Members discussed bringing those capabilities together without overlooking the institutional knowledge and development opportunities already inside the team.

How cash investment roles will change. Looking toward 2030, members of the investing group said their roles are already expanding beyond a narrow portfolio-management remit into cash operations, technology, FX, capital markets, risk management and strategic planning. As forecasting, reporting and eventually parts of execution become increasingly automated, the job may shift toward managing those processes, interpreting outputs, handling exceptions and coordinating with a broader set of stakeholders.

That also creates an interesting talent question. Broad treasury rotations can build the judgment needed to understand how investments interact with liquidity, funding and risk, while increasingly sophisticated technology may create more demand for deeper data and systems expertise. The future treasury team may need both rather than choosing between generalists and specialists.