The New Literacy: What AI Requires of Treasury Teams
If you are feeling behind on AI, you are not alone.
According to the AFP’s survey on AI implementation, 55% of treasury and finance professionals say the number one barrier to AI adoption is a lack of knowledge and competence to work with it. Not budget. Not executive buy-in. Knowledge.
When our Chief Product Officer, Ed Barrie, polled the treasury community on LinkedIn after our first Decoded: AI x Treasury keynote session, the results confirmed it: 20% rated themselves as beginners, 36% said they understand the basics, 20% are comfortable with the concepts, and 24% consider themselves advanced. An almost perfect bell curve. Most of us are somewhere in the middle, aware that AI matters to treasury but unsure how to put it to work.
In this month’s Inside Treasury4 edition, we break down what AI literacy actually means for treasury: how language models produce output, why that output is inconsistent, what you can control, and the one capability that is helping practitioners move from experimentation to repeatable workflows.
These ideas come from the opening keynote of our Decoded: AI x Treasury series, and they apply whether you use Claude, ChatGPT, Gemini, or any other platform.
Also inside: Upcoming events, top treasury roles open now, and more.
Keynote Highlights: The New Literacy in Treasury
The opening keynote session of Decoded: AI x Treasury features Abhilash Nair PhD, who leads the AI and Technology practice at WestCap. Nair oversees the WestCap AI Collective, a working group of engineers and operators across more than 20 portfolio companies, and holds a PhD in Physics. He has been a mentor to Treasury4 since the company’s founding.
We asked him to open this series for a simple reason: Abhilash has a rare ability to teach AI in a way that is both technically grounded and immediately accessible. That is exactly what treasury needs right now.
Understanding what AI can and cannot do
The AFP data shows that knowledge, not budget, is the barrier. Building an accurate mental model of what AI can and cannot do is where literacy begins. Nair opened the session with a framing worth carrying into every vendor evaluation, every pilot, and every conversation about where to invest your team’s time.
“First, you have to believe both of these statements: AI is dumb, and AI is powerful. Both. If you only believe AI is dumb, you’re gonna miss this wave. If you only believe AI is all-powerful, you end up falling for false promises from vendors.”
Abhilash Nair, PhD, AI Practice Lead at WestCap
This matters for treasury specifically because the stakes of trusting unreliable output are high. A hallucinated number in a cash position report is not a minor inconvenience. Understanding where AI is reliable and where it requires verification is foundational, and Nair spends time in the session explaining exactly why models produce inconsistent results. That explanation centers on one concept.
Why output quality depends on context
The most important idea in the session, and arguably in AI literacy broadly, is context. In AI, context refers to everything the model has access to when it generates a response: every message in your session, every uploaded file, every connected tool. All of it is re-sent each time you hit enter. That cumulative input determines the quality of what comes back.
“The model is never the reason why your output is bad. You can take a bad model, give it pristine context, and get a good result. You can get the best model out there and confuse it very easily.”
Abhilash Nair, PhD, AI Practice Lead at WestCap
This explains why the same tool produces a sharp analysis one day and an unusable one the next. The model did not change. The context did. Nair illustrates this in the session with a Plinko analogy that makes the underlying mechanics intuitive.
The practical application for treasury teams: keep each AI session focused on a single task. Mixing topics, or combining planning and execution in the same thread, introduces noise the model treats as signal. When the task changes, start a new session.
From prompt engineering to context engineering
Understanding context also explains why last year’s best practices have already shifted. The standard advice twelve months ago was to build prompt libraries, repeat key instructions, and write with urgency. Nair described that approach as a “hope and prayer strategy.” Prompts are probabilistic. There is no guarantee the model will weight a given instruction the way you intend.
The emerging discipline is context engineering: rather than perfecting the question, you manage what the model can see. That means removing tools you are not using, keeping sessions clean of tangents, and being deliberate about what information is present when the model generates its response.
For treasury, this distinction is practical. The question shifts from “how do I ask AI to build a better cash forecast” to “what should the model have access to when it builds that forecast.” Nair works through this shift in detail during the keynote.
Skills: the capability that brings consistency to AI output
Context explains why output quality varies. Skills are the capability that gives teams control over it. This was the most forward-looking part of the session and the most relevant for treasurers thinking about scaling AI beyond individual use.
Skills are different from tools or connectors. Rather than loading an entire instruction set into the session at the start, skills use progressive disclosure: the model sees a table of contents and only loads the full instructions when the task requires it. That keeps context focused and output more predictable, which is what treasury workflows require.
Nair recommended two starter skills for any business professional. The first teaches the AI to match your writing style, so output reads like your work rather than a machine’s. The second maps your calendar and identifies which meetings are essential, which can be rescheduled, and what follow-up communications are needed.
“If all you take away from this talk is that you should learn how to build a skill, that is going to be a productive exercise for you.”
Abhilash Nair, PhD, AI Practice Lead at WestCap
The session dives deep on how skills work at a technical level, why they are context-efficient, and how organizations can build shared skill libraries across teams. For treasury departments moving beyond experimentation toward repeatable AI workflows, this segment of the keynote deserves close attention.
An honest assessment of what AI adoption requires
Nair was candid about what adopting AI actually looks like in practice.
“You will work more. You will have greater cognitive load. And you will not make more money.”
Abhilash Nair, PhD, AI Practice Lead at WestCap
That is consistent with what most treasury professionals who have started using AI already know. The question is whether the cost of building this competency now is higher than the cost of building it later, while peers and competitors move ahead.
Treasury4’s Chief Product Officer, Ed Barrie, and Lead Data Architect, Cooper Strout, joined for Q&A and shared their experience of going all-in on AI at Treasury4 over the past year. Barrie described the organizational pace as similar to Formula One: periods of acceleration followed by deliberate pauses to let the team absorb what they had learned, then acceleration again. Strout offered a specific benchmark:
“The difference between somebody who is really good at AI versus somebody who’s not is three weeks.”
Cooper Strout, Lead Data Architect at Treasury4
Watch the Full Keynote On-Demand
Wherever you fall on the AI literacy spectrum, this session offers a clear, structured foundation for working with AI more effectively. Nair covers how language models work, how to manage context for better output, and the one capability every treasury professional should learn next.
Treasury4 Tip
Decoded: AI x Treasury
Our virtual session series brings together AI practitioners, treasury leaders, and technologists for candid conversations on what AI means for treasury operations. New sessions drop regularly.
View the SeriesMeet Treasury4 at Upcoming Events
- May 7: San Francisco Symposium
- May 12: Decoded: AI x Treasury Series: Introducing Treasury4’s MCP Server
- May 14: NW Summit | Session: The Three Worlds of Treasury with Christina Easton, Director Treasury Services at Seattle Children’s Hospital and Sunnie Ho, Global Head of Cash Operations at Atlassian
- May 19: Decoded: AI x Treasury Series: The AI Governance and Architecture behind Treasury4
- May 19-20: Windy City Summit | Meet us at Booth 204 | Attend our Sessions
Click here to see the full list of upcoming Treasury4 events.
🚨 Hot Treasury Roles Open Now
- Director Corporate Treasury at Okta | Onsite | San Francisco, CA | Apply Here
- Treasury Lead at Lightning AI | Onsite | NY or San Fran | Apply Here
- Director of Global Treasury, Debt and Capital Markets at Simplot | Onsite | Boise, ID | Apply Here
- VP Fund Accounting, Treasury at Stepstone | Hybrid | San Diego, CA or Charlotte, NC | Apply Here
- VP Tax and Treasury at iRhythm | Remote | Apply Here
- Senior Treasury Manager at ecoATM | Remote | Apply Here
- Assistant Treasurer at Acushnet | Fairhaven, MA | Apply Here
- EVP Finance and Treasury at 4As | Hybrid | New York | Apply Here
- Director of Treasury at Children’s National | Onsite | Silver Spring, MD | Apply Here
- Senior Treasury Manager at Litera | Hybrid | New Jersey | Apply Here
The treasury workforce is changing. The tools are changing faster. AI literacy is no longer a specialization. It is becoming a core competency for practitioners at every level, from analysts building their first forecast to senior leaders evaluating where to invest. The treasurers who build these skills now will define how the next era of treasury operates.
Thanks for reading Inside Treasury4. Follow Treasury4 and Subscribe for more real-world insights, product updates, and treasury industry news.
