By Sagar Shankaran, Founder of CallSphere
Which AI in a franchise group is regulated in 2026, which is genuinely out of scope, and the 18-hour file a multi-unit operator should build this month.
Key takeaways
Which of the AI tools already running inside your fourteen stores has a legal date attached to it, and which ones do not?
Most multi-unit operators cannot answer that, and the reason is not carelessness. You did not buy "AI." You bought Workstream or Harri or Fountain to handle crew applications because you hire 400 people a year and your GMs were losing candidates to the store down the road. You bought a scheduling product that recommends shifts. Your franchisor turned on a guest-feedback tool that writes the first draft of the reply. Somewhere in there, one of those became a regulated use and the others did not, and nobody sent you a letter about it.
Take the three most common ones in a franchise group this size. The crew-application screener that ranks or filters candidates before a GM sees them. The demand forecast that tells your GM to schedule two fewer people on a Tuesday. The AI voice agent or web chat that talks to guests about hours, catering and complaints.
Only the first is the one lawyers are circling. A high-risk AI use, in the way these 2026 laws mostly define it, is one that makes or materially informs a decision about a specific person — hiring, pay, promotion, discipline — as opposed to a decision about product, inventory or hours of operation.
The forecast is a business decision about volume. The guest agent is a conversation. The crew screener decides who gets an interview, and that is an employment decision no matter that it costs you $79 a month and came bundled with your applicant tracking.
Here is where things stand as of late July 2026, and you should read it with your employment counsel rather than from a blog, this one included. Illinois amended its Human Rights Act with AI-in-employment provisions that took effect 1 January 2026. Colorado has an act covering decisions that carry consequences for people, and its start date has already moved once, so confirm where it stands before you build a plan around a specific month. Texas's TRAIGA also took effect 1 January 2026 and turns largely on intent rather than imposing a blanket audit duty on every employer. New York, Utah, Nevada and Maine each have their own statutes; if you operate stores inside New York City there is a separate city rule requiring an independent bias audit and candidate notice for automated employment decision tools, and it has been enforceable for a while now.
Federal preemption of all this is unsettled as of July 2026. That matters practically: you cannot wait for Washington to sort it out, because the state your stores sit in binds you today.
Hear it before you finish reading
Talk to a live CallSphere AI voice agent in your browser — 60 seconds, no signup.
The EU AI Act's high-risk and transparency obligations carry a 2 August 2026 compliance date, and it reaches companies outside Europe whose systems affect people in the EU. Every franchise conference this spring had somebody on stage implying this lands on all of you. For a franchisee with fourteen US stores and US-only crew, it generally does not.
Where it does bite in franchising is narrower and specific. If you are the franchisor rather than the franchisee and you award or screen international franchise candidates, and your candidate scoring runs through an AI tool, you are making decisions about people in Europe. If you run a master franchise or an area development agreement covering EU territory. If your brand's careers portal accepts applications from EU residents for corporate roles and screens them automatically. Those are real. "We ship a bit of merchandise to Ireland" is not.
flowchart TD
A["List every AI tool touching the 14 stores"] --> B{"What does it decide?"}
B -->|"Who gets an interview"| C["Employment decision: notice, human review, records kept"]
B -->|"Talks to a guest as if it were a person"| D["Disclosure: say plainly it is an agent"]
B -->|"How much lettuce to order Tuesday"| E["Out of scope: write one line and move on"]
C --> F["Named owner on the org chart, re-checked every year"]
D --> G["Line in the greeting, the on-hold message and the web chat"]
California SB 53 took effect 1 January 2026 and gets quoted at operators constantly. Read who it is aimed at: the companies building the large frontier models, not a franchisee running fourteen units in Bakersfield. You are not in scope for it as an operator. Do not let a vendor use it to sell you a compliance module.
Also out of scope: your food and labor forecast, your inventory ordering, your drive-thru timer analytics, your co-op marketing spend, your equipment maintenance alerts, and the reader that pulls invoice lines off a produce ticket. None of those decide anything about a person. Write one sentence about each in your register saying what it does and why it is not an employment or guest-facing decision, and you are done with them.
The middle category is disclosure. If a guest is talking to an agent — on the store line, in the web chat on your local pages, in a text thread about a catering order — say so. It is cheap, several regimes point at it, and in my experience guests do not care as long as they are not tricked.
What you are actually being asked to produce is not a compliance program. It is a file that shows you knew what you were running, that a person reviewed the consequential outputs, and that candidates and guests were told. Here is what that costs in hours, as an illustration for a fourteen-unit group.
| Item | Who | Hours |
|---|---|---|
| Register of every AI tool in use, including the ones the franchisor turned on | DO | 4 |
| For the crew screener: vendor's documentation, bias audit if applicable, notice language | DO plus HR admin | 6 |
| Written rule that a GM reviews every screened-out candidate list weekly | DO | 2 |
| Disclosure line added to phone greeting, on-hold message and web chat | Marketing coordinator | 2 |
| Records retention: keep applications, scores and the human decision | Bookkeeper | 3 |
| One hour of employment counsel to read it | Outside counsel | 1 |
| Total | 18 hours |
At a blended $45 an hour internally plus a $400 legal hour, that is roughly $1,165 and one weekend. I cannot honestly price the downside for you — nobody can price a discrimination claim in advance — but I can tell you which side of that trade every franchise attorney I have talked to this year lands on.
Three things do not belong on a template, and this is the honest part.
Still reading? Stop comparing — try CallSphere live.
CallSphere ships complete AI voice agents per industry — 14 tools for healthcare, 10 agents for real estate, 4 specialists for salons. See how it actually handles a call before you book a demo.
The first is the question of who is responsible when your franchisor mandated the tool. Many brands now require a specific applicant tracking product as an approved supplier. You are still the employer of record — you sign the I-9s, you are on the wage claim — but the tool was not your choice. Whether your franchise agreement's indemnity clause reaches this is a real question, and it is worth one paid hour to have your franchise counsel read the technology and indemnification sections together. Do not raise it with your franchise business consultant as an accusation; raise it as a question about approved suppliers, which is a conversation brands are having with a lot of operators right now.
The second is multi-state exposure. If your fourteen units straddle Illinois and Indiana, the crew screener is doing two different things legally at the same time. The cheap answer most operators take is to run the strictest state's practice everywhere — notice, human review, records — rather than maintain two processes at store level.
The third is that none of this substitutes for actually reading the screened-out list. A file that says a human reviews it, attached to a human who does not, is worse than no file. Put it on a named person, on a named day, in the same weekly rhythm as the labor recap.
For their systems, possibly. For your employment decisions, no. In almost every US franchise structure the franchisee is the employer, and a brand letter saying a tool was vetted is not a defense to a claim brought against your entity. Keep your own copy of whatever documentation the brand provides and keep your own record of the human review.
Say it is an assistant in the greeting. It takes four words, it satisfies the disclosure direction all of these regimes are pointing, and it costs you nothing operationally. Where it becomes a live legal question rather than good manners is recording — call recording consent is a separate body of state law that predates all of this, and two-party consent states still apply.
Write the register anyway. It is an hour, and the value is that when the franchisor, an insurer or a plaintiff's lawyer asks what you run, you have an answer that took a weekend instead of a scramble. Scheduling deserves a second look, though: if the tool assigns individual shifts to named employees rather than recommending headcount, ask counsel whether that crosses into an employment decision in your states.
No. Preemption is unsettled as of July 2026, and in the meantime the state statutes are the ones with effective dates that have already passed. Waiting is a choice to be non-compliant in Illinois today in exchange for a federal outcome nobody can date.
Block two hours and write the register: every tool, what it decides, which of the three buckets it falls into. Ninety percent of a fourteen-store group's list will land in "out of scope, one line." The one or two that do not are where your weekend goes, and after that it is an annual re-read, not a program.
If part of your list is the phone and chat that talk to guests, the disclosure piece is small and worth doing properly. CallSphere builds AI voice and chat agents for business phone lines and web chat that answer around the clock, book appointments and capture leads — and identifying themselves as an assistant in the greeting is a setting, not a project. Get that language agreed with your counsel once and use it across all fourteen store numbers.

Written by
Sagar Shankaran· Founder, CallSphere
LinkedInSagar Shankaran is the founder of CallSphere, where he builds production AI voice and chat agents deployed across healthcare, hospitality, real estate, and home services. He writes about agentic AI, LLM engineering, and shipping voice agents that handle real calls in production.
See how AI voice agents work for your industry. Live demo available -- no signup required.
The EU AI Act's August 2 date, Texas TRAIGA and California SB 53 all landed. What a US rehab clinic must document, disclose and log, and what it can skip.
Handwritten shorts on back-door delivery tickets leak thousands a year across a 14-store group. What 2026 document reading changes, with the arithmetic.
Driver screening tools and cab-facing cameras put carriers under Illinois, Texas and NYC rules in 2026. What to document, and what is genuinely out of scope.
Two casino systems land in the EU AI Act high-risk bucket: face matching and marker scoring. What US gaming operators document, disclose and can ignore.
Why multi-unit franchise operators sampled the exception report, what the 2026 collapse in AI cost changes, and a worked example on 4.6 million tickets a year.
What a US registered investment adviser must actually document, disclose and log in 2026 — and which AI statutes are genuinely out of scope for the firm.
© 2026 CallSphere Inc. All rights reserved.
Made within San Francisco
Watch how CallSphere handles real customer calls, schedules appointments, and processes payments — live.
Try Live DemoBook a DemoCalculate Your ROI