Bank branch of 2026: the technology stack that actually moves the P&L.

The five-year decline in U.S. branch counts inverted in the first quarter of 2026. FDIC data shows 267 branches opened and 217 closed between January and March 2026 — the first net-positive quarter in more than a decade. But the new-format branch is not the four-teller-window box that closed. Mid-market community and regional banks buying branch technology this year are buying a smaller footprint built around three moves: interactive teller machines and cash recyclers that push cost per teller transaction down 40–70 percent; a universal-banker workstation that collapses teller, CSR, and light-lending specialization into one role; and video banking that lets a regional network staff any branch as if it had a mortgage specialist on site. Each move is available from a shortlist of named vendors. Each has failure modes that show up in the same three or four places every sourcing cycle.

The 1Q 2026 branch-count reversal is real, but the branch that is opening is not the branch that closed.

The FDIC Quarterly Banking Profile for the first quarter of 2026 recorded 267 branches opened against 217 closed — a net-positive quarter for the first time since before 2015. Community bank net interest margin held at 3.71 percent in the same quarter, restoring the deposit-side economics that make new physical distribution defensible. The structural point for a sourcing lead is that the new openings are smaller-format, higher-throughput, and technology-dense. Full-service branches with four teller windows and a call-in loan officer are still closing. The openings are 1,800–2,500 square-foot lobby formats staffed by two or three universal bankers with an ITM or two, a cash recycler behind the counter, and a video connection to specialist bankers who sit centrally. That is a different procurement problem than replacing a 1990s tellered branch on the same footprint. The authoritative branch dataset for anyone modeling their own footprint is the FDIC BankFind Suite Summary of Deposits, which publishes branch-office records for every FDIC-insured institution with historical filings.

The “NCR” on your RFP is now two companies with different roadmaps.

In October 2023 the legacy NCR Corporation split into NCR Voyix, which holds retail POS and enterprise software, and NCR Atleos, which holds the ATM and interactive teller machine line plus the associated managed-services book. Every ITM and cash-dispensing contract signed with the banking side of the pre-split NCR now sits with Atleos. Contracts inadvertently paper-named against Voyix have started to surface at renewal, and the roadmap question for banking hardware is unambiguously an Atleos question. The vendor primary on the ITM category — the Atleos explainer on interactive teller machines — makes the boundary clear: video-collaboration and remote-transaction processing are embedded on Atleos-branded hardware, not on anything Voyix ships. Any RFP that names only “NCR” without a corporate qualifier should be corrected on the way to signature. This is the single most common contract hygiene error observable in the 2026 cycle.

Interactive teller machines pencil hardest in branches doing 150–400 transactions per day.

The three ITM hardware lines in serious contention for a mid-market bank in 2026 are NCR Atleos's 80 Series, Diebold Nixdorf's DN Series with the Vynamic Branch Transformation software layer, and Hyosung Americas's MoniMax family (the 5600 for lobby, the 7600T for higher throughput) running the ActiveTeller software layer for the video-teller connection. All three combine cash dispense, deposit automation, and interactive-teller video into a single device. The category diverges on managed-services economics — Diebold Nixdorf and NCR Atleos both leaned harder into managed services in 2025 and 2026, and Atleos's post-split independence has accelerated its managed-services push — and on the depth of production integration with each core banking platform. Deployment economics do not depend on the hardware datasheet. They depend on transaction density. Below roughly 150 transactions per day, an ITM sits idle enough that the payback breaks. Above roughly 400, throughput bottlenecks appear and a second device or a hybrid tellered configuration is warranted. The band in between is where the case is easiest to make.

Cash recyclers still pencil, because cash usage did not go to zero.

The case for branch cash automation is not a bet on cash growth. It is a teller-hour recovery case anchored to a real usage baseline. The Federal Reserve's 2026 Diary of Consumer Payment Choice confirms that four in five consumers used cash in the prior 30 days — the number cited across Hyosung and Glory vendor material and consistent with the Fed's payments-research posture. The active vendor names in cash automation are Glory Global Solutions, Cummins Allison (now part of Crane Payment Innovations after the February 2020 acquisition — a corporate-identity detail that still misfires in RFPs that assume a Loomis relationship), and Loomis on the cash-in-transit and cash-management side. Vendor claims of 30–50 percent teller-hour reduction are the negotiable variable and the appropriate anchor for a contract. Buying the device count without contracting to the teller-hour number is buying the wrong thing.

Video banking has consolidated faster than most buyer shortlists reflect.

Two acquisitions reshaped the video banking category between 2022 and 2023, and most 2026 RFP shortlists still read as if neither happened. Eltropy acquired POPi/o in June 2022 and has since folded video into a Unified Conversations platform covering text, secure chat, eSign, eNotary, and video check deposit; the current Eltropy footprint is more than 400 credit unions and a rising set of community banks. Glia acquired Finn AI in June 2023, adding a virtual-assistant layer to its digital customer service platform, and distributes through Jack Henry. Coconut Software runs the third serious position, pairing appointment scheduling with video and, since the June 10, 2026 launch of AI-powered Branch Workforce Management, staffing analytics — Coconut cites a demand-forecast error rate under 7.5 percent against a 15–20 percent industry range in its own launch materials. LiveBank from Ailleron shows up in some U.S. deployments and should be verified at contract stage per client. The pattern buyers repeatedly miss: video banking without appointment scheduling upstream is a call center with a camera in a branch, and utilization stays under any reasonable threshold. Eltropy or Glia paired with Coconut (or an equivalent scheduling integration) is a live pattern; either one alone is not.

The universal banker workstation only pencils if the core exposes the right APIs.

“Universal banker” is a staffing model — one employee who can open an account, take a deposit, initiate a consumer loan application, and handle basic servicing — that requires a workstation the core supports. On the Fiserv side the relevant products are Connected Teller for Signature, Teller for Cleartouch, and tMagic for DNA; on the Jack Henry side, the branch modules that sit on Symitar and SilverLake; on the FIS side, the branch banking modules that sit on Horizon and IBS. A workstation buy that does not confirm production integrations against the specific core deployment ends up in a multi-quarter integration project that eats the payback the branch-format change was supposed to deliver. The sequencing rule is straightforward: the universal-banker workstation should align to the next core-refresh window, not lead it. Buying a workstation ahead of a core migration commits to integration debt that gets written off when the core moves.

Appointment scheduling pairs with branch analytics — but only if the FI acts on the schedule output.

Appointment scheduling is the highest-ROI branch technology purchase available in 2026 and is also the one most frequently bought and then not operationalized. Coconut Software runs the deepest bank- and credit-union-specific product footprint, with more than 200 FI clients including RBC, M&T Bank, and Mountain America CU per its own product materials. Q2 and Alkami both bundle scheduling as part of their digital banking overlay; integration depth against an incumbent Coconut relationship is worth verifying rather than assumed. The Coconut launch of Branch Workforce Management on June 10, 2026 (bWFM, per PR Newswire) added AI-driven demand forecasting and shifted the category from “scheduling with light reporting” to “scheduling plus staffing optimization.” The value is only usable if the branch operations team actually reroutes staffing based on the forecast. Left ignored, the forecast is a screen no one opens. Coalition Greenwich's small- and middle-market business banking coverage — the 2026 U.S. Middle Market and Small Business Banking Best Bank studies — continues to show branch-driven satisfaction premiums in the middle market segment; the technology only shows up in those numbers if the physical branch is staffed correctly to convert the appointment.

Sequence the buy across 24 months, and negotiate managed-services alternatives where hardware ownership breaks.

The sequencing rule that survives contact with the actual budget cycle is a two-year plan. Year one: cash recyclers in the highest-density branches, ITMs in three to five pilot branches that clear the 150-daily-transaction threshold, and appointment scheduling network-wide. Year two: video banking integrated with the scheduling upstream and branded inside the mobile channel, plus universal-banker workstation aligned to the next core-refresh window. Branch analytics — meaning the workforce-management output layer — should wait until at least twelve months of scheduling data exist to inform it. Both Diebold Nixdorf and NCR Atleos pushed managed-services alternatives harder in 2025 and 2026, and for banks under roughly 30 branches the total-cost math frequently flips against outright hardware ownership. That is a contract structure to price out on any hardware refresh, not an afterthought.

Run the sourcing decision framework before writing the RFP.

The nine-item framework below is the shortest usable pre-RFP work product for a $2B–$20B community or regional bank sizing a 2026 branch-technology procurement. It is the artifact this piece exists to hand to a chief operating officer or head of retail.

What breaks: three failure modes visible in every 2026 mid-market branch sourcing cycle.

The RFP names “NCR” without specifying Atleos or Voyix. Contracts have been executed against the wrong entity since the October 2023 split. ITM and ATM roadmaps sit with Atleos; retail POS sits with Voyix. Verifying the counter-party in the master services agreement takes an hour and prevents a multi-year contract mis-anchor.

ITM utilization craters in low-density branches. A bank buying ITMs to remove teller headcount discovers the device sits idle 60 percent of the day in branches under 100 daily transactions. The remediation is either a density-based deployment plan on the front end or a redeployment plan on the back end; the choice needs to be made before the hardware arrives.

Video banking video quality tanks in older branches. The software is not the constraint. Bandwidth to the branch and endpoint audio-visual quality in the room are. Vendors deliver working platforms into unusable branch AV rooms, and the branch-side team blames the vendor. Any video banking rollout needs a branch AV assessment on the front end.

What this means for procurement in 2026.

The 2026 branch procurement discipline is narrow and executable. Confirm the corporate identity on every ATM and ITM contract. Anchor the ITM business case to real transaction density, not to a headline reduction claim. Contract cash recycler procurement to teller-hours, not devices. Pair every video banking purchase with an appointment scheduling integration on the front end and a branch AV assessment behind it. Sequence the universal-banker workstation to the core-refresh window rather than ahead of it. Price out managed services against capex for any bank under 30 branches. The branch is not the branch of 2015. The stack is not the stack of 2015. A sourcing plan written against the old assumptions will not clear the current audit or the current P&L expectation.

This is the fourth piece in the Financial Services Analyst Note series. The anchor is How mid-market financial services operators should source technology contracts in 2026. Related pieces: PCI DSS 4.0 for mid-market merchants and FFIEC IT booklet mapping for mid-market banks. Every vendor named here is in The Cardinal Source's active supplier pool.

Sources

• FDIC, Quarterly Banking Profile — First Quarter 2026. fdic.gov
• FDIC news release, “FDIC Quarterly Banking Profile First Quarter 2026.” fdic.gov
• FDIC BankFind Suite, Summary of Deposits — branch office data. banks.data.fdic.gov
• FDIC BankFind Suite, bulk data download. banks.data.fdic.gov
• NCR Atleos, “What is an interactive teller machine?” ncratleos.com
• Diebold Nixdorf, “Branch Automation Solutions” portfolio release (2025). prnewswire.com
• Diebold Nixdorf, Vynamic Branch Transformation product page. dieboldnixdorf.com
• Hyosung Americas, benefits of interactive teller machines. hyosungamericas.com
• Fiserv, Connected Teller for Signature. fiserv.com
• Fiserv, tMagic for DNA. fiserv.com
• Eltropy, “Video Banking for Credit Unions and Community Banks Enters a New Era with Eltropy.” eltropy.com
• Glia, “Glia acquires Finn AI.” glia.com
• Coconut Software, “Coconut Software Launches AI-Powered Branch Workforce Management Solution” (June 2026). prnewswire.com
• Coconut Software, product overview. coconutsoftware.com
• Coalition Greenwich, corporate banking coverage. greenwich.com

All linked sources were live at time of publish (July 2026). Verify before quoting in a procurement document.