Blog

Beyond prompts to playbooks: what TAI Skills mean for modern treasurers

Most treasurers using AI today are using it the same way: ask a question, get a faster answer. That's useful. It saves time. It cuts down on screen-hopping.

What if AI could do more than answer questions? What if it could run the work for you? Multi-step, complex, on demand, the way you actually do it. With a record of every step, ready for you to validate.

That's what Skills in TAI (Kyriba’s Trusted Agentic AI) are designed for.

What is a Skill in TAI

A Skill is a reusable set of instructions that teaches TAI how to perform a specific treasury task consistently. You define the process once, then run it whenever needed.

Put simply, a Skill turns treasury know-how into an executable playbook.

That playbook might capture a recurring cash forecasting analysis, a periodic treasury briefing, a daily liquidity monitor, or any other repeatable process that depends on connected steps, judgement, and context. Instead of rewriting prompts each time, or relying on someone to remember the exact sequence of actions, the Skill gives TAI the procedural knowledge it needs to produce a consistent outcome on demand.

This same shift is happening across the AI world, from Google, Claude, Microsoft. The reason is simple: a one-time prompt is helpful, but a repeatable practice is transformational. In treasury, where the same complex tasks repeat across entities, currencies, banks, and counterparties, that difference matters even more.

So how does TAI actually do this? TAI starts with a frontier AI model that already has deep knowledge of finance. On top of that, we've added everything Kyriba has learned about treasury over more than 20 years working with our customers: the practices, the patterns, the judgment calls that make treasury work in the real world.

Skills take this one step further. They let you tailor TAI to your company with your operating procedures, your policies, and the specifics of how your team actually runs treasury.

And here's what's new: this customization doesn't require programming, scripts, or configuration menus. You write it in plain English (or your local language), the same way you'd explain the task to a new colleague.

English (or your local language) is the new programming language.

That's why a Skill does not just help you get to an answer faster. It helps you standardize how the work gets done — your way.

The value you discover with TAI

The treasurers who get the most from TAI are the ones who keep discovering. Most start with quick answers. The bigger shift happens when they realize how much more is possible with TAI, especially with the introduction of the Skills capability. Let's take a closer look.

A diagram showing the value customers can discover with TAI, Kyriba’s trusted agentic AI. Most start with quick answers to gain productivity. The bigger shift happens when they move forward to accuracy, control and intelligence.

1. Quick Answers

This is the entry point. It builds familiarity with what AI can do, and it builds trust through transparent, sourced responses. Most treasurers begin here. It's the right place to start. But many also stop here, leaving the bigger value untouched.

The use case: Use TAI to query your data and get fast answers, the kind of answers you'd otherwise look for across multiple screens or reports. You can also ask TAI how to use Kyriba itself, and get a sourced answer in seconds, instead of digging through documentation or waiting on support.

The value: the work you'd do anyway, minus the time it usually takes. Productivity is the first and most visible gain, and the most important one for many treasurers when they start with TAI. It's also the level where trust is built and curiosity grows about what else is possible.

Example queries:

  • "What's my cash position across all entities right now?"

  • "Which payments are awaiting my approval?"

  • "What's the variance between today's actuals and forecast?"

  • "How do I create a payment template in Kyriba?"

2. More Precision

The shift from getting answers to improving the underlying quality of the work you already do every day: cleaner, sharper, and more reliably. The places where small data quality gaps could quietly compound into bigger downstream problems.

The use case: Use more complicated queries for insights or start to build Skills in TAI for precision.

Example Skill: Unknown Cash Flow Identification

The pain: Cash flow forecasting only works if every flow is correctly classified. But in real treasury operations, transactions arrive at every entity, every day, often without proper budget codes. Treasurers either guess, chase the source, or quietly accept inaccuracy.

The cost of "unknown" cash flows isn't messy reports. It's eroded forecast credibility with the CFO and stakeholders.

What the Skill does: It scans recent cash flows, identifies those without proper classification, applies pattern-matching against historical data, and recommends correct coding. Ambiguous cases are routed for human review.

Why it matters: Forecast accuracy is the cornerstone of liquidity performance, and the silent killer when it slips. This Skill closes a gap no dashboard can show you. Every forecast cycle starts cleaner. Every variance conversation gets sharper. A Skill that systematically tightens cash flow data quality systematically improves the credibility of every forecast that follows.

/identify_cash_flows*

  • Pull list of cashflows that are mapped to proper budget code

Step 1 : pull a list of cashflows with budget code UNIDENT for this month

Step 2: pull a list of available budget codes from core data

Step 3: Analyze common budget codes for the accounts of the unidentified transactions

Step 4: Analyze cash flows descriptions and characteristics, cross reference with characteristics of unidentified transactions and budget code trends

Step 5: Provide a table of unidentified transactions with a column listing the suggested budget code. provide explanation of reasoning, included in a table.

3. More Depth

This is where Skills bring rigor to the work you do, but rarely as thoroughly or as often as you'd like. The kind of practice your team knows is valuable, but defers because the manual effort is high. With Skills, that depth becomes available on demand.

The use case: Use Skills to add depth and consistency to the work you do, run it whenever needed with minimum effort required.

Example Skill: Pool Finance Setup Watchdog

The pain: Working capital finance — supply chain finance, receivables finance, dynamic discounting — depends on dozens of configurations across counterparties, programs, limits, and policies. A single misconfiguration can create financial exposure, compliance risk, or unexpected cost.

Today, those checks happen manually, or not at all. They're complex, high-risk, and easy to overlook.

What the Skill does: It performs an end-to-end review of pool finance setups on demand, validating configurations against policy and best-practice patterns, and surfacing anomalies for review before they become problems.

Why it matters: Working capital finance is high-impact and high-risk without data integrity in place. A Skill that polices its setup at the configuration level protects margin and reputation simultaneously. Run it before a new program goes live. Run it as part of a periodic review. Run it any time the stakes feel high. This Skill lowers the operating cost of complexity — the kind of practice where one undetected error costs more than anyone could predict.

/pool-setup-watchdog*


  • Run the “TAI Audit Trail” (TAI_AUDITTRAIL) report and analyze the data for the last 24 hours. Check for any Creation, Modification, or Deletion across the following entity types. Organize results by impact tier and apply all flagging rules below.

---

  • Data integrity rules – apply before any analysis:

  • If data is missing, state it clearly.

  • Do not proceed with partial data. Ask the user to confirm any ambiguous inputs first.

---

  • Entity list to check:
    • Tier 1 (direct impact on calculation):

    • RF - Deal Parameter | SC - Early payment contract | SC - Early payment rule | SC - Interest term | FT - Credit line | FT - Limit

    • Tier 2 (affects eligible amounts):

    • RF - Drawdown request | RF - Reserves | RF - Unallocated cash | SC - Documents

    • Tier 3 (program configuration):

    • SC - Supplier service access | SC - Processing option | SU - Third party | SU - Company

---

  • Flagging rules: …

---

(The Skill shown above is a simplified excerpt.)

4. New Strategies

At this level, the shift is from doing what's possible today to doing what isn’t possible yet. The valuable practices everyone agrees with, but no one finds time or budget for. This is where Skills change the game the most.

The use case: Use Skills to do the work that brings great value across the organization but isn’t done because it's too time-consuming, too easy to defer, or because the data is too fragmented.

Example Skill: Treasury Efficiency Audit

The pain: Every treasury operation drifts over time. Reconciliation gaps creep in. Manual exceptions accumulate. Configurations age out of best practice. Most teams aren't actively checking how their treasury operations are doing. Too busy. No time. Everything looks fine, the data is everywhere. The result is an operation that feels fine, until the day it doesn't.

The challenge has always been the same: how do you audit operational efficiency rigorously, without the cost of a full audit project?

What the Skill does: It runs a multi-dimensional review across treasury operations, flagging inefficiencies, surfacing exceptions, identifying configuration drift, and producing a prioritized recommendations report with clear rationale.

Why it matters: This is the kind of analysis that would normally take a consultant days. With a Skill, you can run it whenever you need it. Catching drift early, demonstrating governance maturity, surfacing savings opportunities you'd otherwise miss. A Skill that audits the operation rigorously is a Skill that protects it rigorously. This is the discipline of continuous oversight.

/treasury_efficiency_audit*

  • Objective: Analyze treasury management efficiency for [Region] over the [Analysis period].

  • Scope: All bank accounts held in [Country], denominated in [Primary Currency].

Step 1 — Pull the data …

Step 2 — Calculate overdraft costs …

Step 3 — Calculate opportunity cost on idle positive balances …

Step 4A — Forecasting gap analysis …

Step 4B — Link forecasting gaps to financial impact …

Step 5 — Account balancing analysis …

Step 6 — Summarize results

  • Table 1 — Cost summary by bank

  • Table 2 — Unforecasted transactions (ORIGIN FIELD starts with "BK/")

  • Table 3 — Companies with weakest forecasting (most "bank" origin transactions)

  • Table 4 — Account balancing gaps (overdrafts vs. idle balances on the same day)

Step 7 — Recommendations based on findings …

(The Skill shown above is a simplified excerpt.)

The shift this represents

Look across what you discover with TAI, and a pattern emerges.

The first thing you find is productivity gain with time saved. The next things you find are bigger:

  • Accuracy — in your data, your classifications, your configurations

  • Control — with full validation, auditable trails, and your judgment in every loop

  • Intelligence — surfacing what matters, with reasoning you can defend

This is Agentic Finance in practice. It brings a depth of rigor to treasury that transforms the way value is created.

The treasurers getting the most from TAI today aren't the ones asking the most questions. They're the ones turning their best practices into Skills, and running those Skills whenever the moment calls for it.

From prompt to playbook. That's the shift.

Take a TAI Product Tour, or talk to us for a demo if you want to learn more.

*All Skill examples in this blog were written by treasury practitioners in plain English. Some have been shortened here for readability. Real Skills written by treasury teams typically run longer and include additional steps, validation rules, and output formatting.

Written By

Félix Grévy

Félix Grévy

SVP Platform, Data & AI

Félix Grévy is SVP of Platform, Data & AI at Kyriba, where he leads innovation across platform engineering, data, AI, and advanced analytics. With more than 20 years of experience in financial technology spanning product development, product management, and commercial management, Félix joined Kyriba in 2020 to lead API and connectivity strategy. He has since spearheaded Kyriba's agentic AI initiatives, including the Trusted AI (TAI) portfolio, which embeds governed intelligence directly into treasury and finance workflows by integrating LLMs and predictive analytics, without "black boxes" or training external models on customer data.

Related resources

Blog

Why AI’s real value for CFOs isn’t just speed, but trusted governance at scale

Read
eBook

How AI agents actually work: A practical guide for Treasury & Finance

Learn more
Blog

Agentic AI is a gold rush for CFOs and treasurers—not a sugar rush

Read