Stop Asking If You Can Afford AI. Start Asking Where the Money Already Is.

Every budget season, the same thing happens. Someone brings an AI proposal to the leadership table, and the conversation immediately becomes a debate about new spend. New line item. New vendor. New thing to defend in front of the CFO and maybe the board.

That framing is why good projects die.

In our latest webinar, Kathy Sianis, SVP of Client Success and Partnerships at Posh, made the case that the question is backwards. She was joined by Steve Goodwine, SVP and Head of Contact Center at Hudson Valley Credit Union, who has now built four AI business cases and gotten them funded.

Here is what they covered, and what you can take into your own budget conversation this week.

The reframe: you are already paying for the problem

New line items require you to defend additional spend. Reallocations do not.

"Start with a business problem," Kathy said. "If the problem already exists, there's almost always money already being spent today."

Sometimes that money is obvious: a vendor contract, an overflow provider, an after-hours service. Sometimes it is buried: SME time, manager coaching hours, travel to training, the 10 minutes every employee spends hunting for an answer they should have found in seconds.

The question is not "can we afford AI." It is "where are we already spending money solving this problem, and is that spend working?"

The framework: five steps, in order

Kathy walked through a repeatable sequence that works for any solution, from any vendor:

1. Business challenge. Too many calls. Training takes too long. Employees cannot find information. Start here, not with the technology.
2. Current process. Who touches it, how long it takes, where work gets repeated.
3. Current cost. Labor, vendors, technology, lost productivity, member impact.
4. Desired outcome. Be specific. Fewer calls? Faster onboarding? Better QA visibility? More self-service?
5. Potential solution. Last, not first.

"Don't start with, we need a chatbot. We need AI," Kathy said. "You really want to start with the business problem."

Five ways AI creates value

Four are the usual suspects. The fifth came from Steve.

• Reduce existing costs. What can you stop paying for, consolidate, or retire? This is the easiest value for finance to recognize.
• Avoid future costs. Growth does not have to mean more headcount, more vendors, or more facilities.
• Improve employee capacity. Giving someone 10 minutes back only matters if you know what they do with it. Tie the time back to higher value work.
• Improve member experience. Faster answers, better availability, fewer transfers, more consistent responses. This shows up in retention, loyalty, and NPS.
• Increase revenue. Steve's entire Training Simulator business case is built on this one.

What Hudson Valley actually did

Voice Assistant: funded by a headcount increase they never made.

Hudson Valley had budgeted for additional headcount in the same year they planned to deploy Voice, driven by two mergers. So they delayed the hire, launched Voice, and never made the increase.

"We were cost neutral to what was budgeted for that year," Steve said.

The first-year goal was modest: match the containment of the touch-tone menu, then improve it by roughly 10% at the 12-month mark. They hit the 10% at go-live. The following year, they reduced headcount by another 1.5 FTE and beat their pro forma.

Knowledge Assistant: the business case was wrong, and it still worked.

They built the case entirely on reduced handle time. At the six-month look-back, handle time had not moved at all.

What happened instead: agents reinvested the time they saved into selling.

"The calls remained the same length, they just became more valuable to the organization," Steve said.

The financial goals were met, just through a different door. New hires reached proficiency faster and started generating revenue three months earlier than before. First contact resolution rose 3%, which for Hudson Valley is the equivalent of 1,500 calls avoided, or about 1.5 FTE. Calls into support and operations departments dropped roughly 35% across the board, freeing those teams to work their own backlogs.

Digital Assistant: 6x the volume, the same staff.

After an internet banking platform change put chat behind authentication, chat volume increased 600%. Hudson Valley now handles about 13,000 live chats a month with the same roughly 1.5 FTE per day they used before the conversion.

The found money nobody was looking for.

While working through intelligent routing, Steve's team discovered they were already paying their card processor to take 24/7 fraud and dispute calls. They had been handling those calls in house anyway, plus paying IT costs to build the forms and workflows to support them. Now those calls route straight to the processor.

That is a cost save that existed before any technology was purchased. It just took the exercise to find it.

The governance bar: cost neutral in year one, 100% ROI in year two

Hudson Valley recently stood up an AI strategy and governance team. Those two numbers are now the minimum for any AI deployment.

That discipline is why the Posh Simulator case is built on loan dollars, not vibes. Steve's team calculated how much incremental lending the tool needs to generate each year to clear the bar, accounted for loan servicing expense and provision for loan losses, and made that number a deployment goal.

Then they left the soft savings out on purpose. Role play requires two people. A simulator requires one, and nobody gets pulled off the phones, which protects speed to answer and abandon rate. Many institutions also pay a vendor extra for a sandbox or training version of their platform, a cost a screen-pathing simulator can displace.

"There are those other lingering effects that we'll expect to see, that we're underselling, right, so we can over-deliver," Steve said.

The 99.7% nobody is listening to

Between his own team and the independent QA department, Hudson Valley reviews three tenths of one percent of calls.

That is not a Hudson Valley problem. It is everyone's problem. Humans can only listen to so many calls.

"We've reviewed more calls isn't the success," Kathy said. "The success is what changed because we had more visibility."

Sales opportunities, first call resolution drivers, fee reversals and courtesy credits caused by bad information, regulatory exposure, empathy and listening gaps that show up in NPS. All of it sits in the calls nobody hears.

Build a portfolio, not a purchase

You do not need everything at once. Start where the business case is strongest and the measurement is clean. Deliver. Then use that credibility to fund the next one.

"I already built up some goodwill through the knowledge deployment," Steve said. "We set a measurable mark, we exceeded that mark."

He also pointed out that the solutions feed each other. CoachQA surfaces the gaps that become Training Simulator scenarios. Posh Outreach gives Voice a way to follow up.

That last one produced the best example of the session. New reporting showed Hudson Valley that 34% of members who try to authenticate into the Voice Assistant fail and get transferred to a live person. Those members already want to self-serve. They just need their member number or their PIN.

"These people have already demonstrated they're willing to engage with a virtual assistant," Steve said. "It's low-hanging fruit for us. We just have to find the capacity to do it."

Six steps to a business case your CFO will fund

1. Current cost. A baseline finance already believes.
2. Future state. Be clear about what actually changes.
3. Investment. Implementation plus recurring.
4. Annual benefit. Separate hard savings from soft capacity gains.
5. Payback. Months to recover.
6. Multi-year ROI. Steve uses a five-year pro forma. Kathy recommends one, two, and three year views.

Anchor every step in a number your finance team already recognizes. That is what turns a proposal into a funded project.

Run your own numbers

We built an AI Business Case Calculator so you can do this without a spreadsheet from scratch. Plug in your call volume, your fully burdened hourly cost, your QA coverage, your training spend, and see what you could reallocate.

Try the AI Business Case Calculator

Want a second set of eyes on it? Kathy runs 30-minute ROI workshops and budget planning sessions with banks and credit unions every week. Not a sales call. Reach out at info@posh.ai.

Plus, Posh Conversations 2027 is April 13 to 15 at The Westin Boston Seaport District. Registration is open.