Tag: small business

On May 13, 2026, Microsoft moved a capability out of preview that quietly changes what “automation” means for small businesses. Computer-use agents in Copilot Studio reached general availability across all commercial Power Platform regions — and unlike the integrations and chatbots that came before, these agents don’t need an API to work. A computer-use agent gets the same tools a human employee gets: a browser, a screen, a keyboard, and the ability to read what’s on the page and take the next step using vision and reasoning. In plain terms, it can operate the clunky legacy software, vendor portals, and web forms that never had a clean connector — the exact systems that trap a small business’s most tedious work.

That distinction matters more than it sounds. For years, “automating” a workflow meant your software had to offer an integration, or you paid a developer to build one. The result: the messiest, most manual tasks — the ones that involve logging into a supplier’s ancient portal, copying numbers between two systems that don’t talk, or re-keying orders from a PDF — stayed stubbornly human because no API existed to hook into. An agent that can see and click navigates those systems the way a person would, which means the automation finally reaches the work that was previously off-limits.

The capability ships with guardrails aimed at exactly the concerns an operator should have. According to Microsoft, the agents run on Windows 365 Cloud PCs, all screen-data processing stays inside the customer’s tenant boundary, and sensitive inputs can be masked using Azure AI Content Safety. Administrators can set allow-lists for which websites and desktop applications an agent may touch, and the usual Power Platform governance — data-loss-prevention policies, environment isolation, audit trails — applies. It’s available to any commercial customer with a Copilot Studio license that includes agent runtime capacity, with government clouds following in the second half of 2026.

For a small-business operator, the honest framing is this: a powerful capability that big companies will use to automate legacy processes is now, in principle, available to you too — and the smart response is selective, not breathless. Start where the work is repetitive, rule-bound, and currently un-automatable because the system has no integration. Re-keying invoices from a vendor portal into your accounting software. Pulling order data from a supplier site that never offered an export. Updating records across two tools that refuse to sync. These are the high-volume, low-judgment tasks where an agent that operates the screen earns its keep, and where a mistake is cheap to catch on review.

The discipline that separates a time-saver from a liability is verification and scope. An agent clicking through your live systems can do real damage if it misreads a screen and submits the wrong thing — and unlike a bad email, a wrong entry in a financial or inventory system propagates. Operators getting this right keep the agent on a tight allow-list, set dollar or volume thresholds above which a human must approve, never let it touch anything legally or financially binding without sign-off, and spot-check its work on a fixed rhythm. Treat it as a fast, tireless assistant whose output you still scan — not a system you switch on and stop watching. The cost of an unnoticed error in the back office is almost always higher than the time the automation saved.

If you’d rather not navigate the setup, governance, and tool selection alone, LevelUpLabs.co is built for operators in exactly this spot: practical AI strategy guides, a prompt library for admin and operations tasks, video training that walks through real configurations, rollout checklists for deploying automation safely, and partner discounts on the underlying tools. It’s a tested path to modernizing the back office without a consultant’s invoice or weeks of guessing what to automate first.

The bottom line for small-business operators: the most tedious corner of your week — the manual data shuffling between systems that were never designed to cooperate — just became automatable, because AI can now operate software the way a person does. That’s genuinely new, and it’s worth acting on. But the winners won’t be the operators who turn an agent loose on everything. They’ll be the ones who pick one un-automatable, repetitive task, fence the agent in tightly, keep a verification step, and measure the hours reclaimed before expanding to the next. Quietly handing your busywork to software, carefully, is how yo

The U.S. Census Bureau released a data story in May 2026 with a blunt title: “Large Firms With at Least 20 Employees Biggest AI Users.” Drawn from the Bureau’s Business Trends and Outlook Survey (BTOS), it found that between December 2025 and May 2026, AI use rose among firms with at least 20 employees but didn’t change significantly among firms with fewer than 20. Among the very smallest businesses — those with four or fewer employees — less than 20% reported using AI at all.

The gap is sharper when you weight by employment. About 18% of U.S. firms reported using AI in at least one business function, but on an employment-weighted basis adoption was 32% — meaning the firms that employ the most people are adopting fastest. Overall AI usage across all businesses hovered between 17% and 20%, with another 20–23% expecting to adopt within six months. There’s also wide sector variation: as of early May, the Information sector was at 39.7% and Finance and Insurance at 33.9%, both well above the national rate of roughly 19.8%.

Before every small-firm owner takes that as bad news, here’s the important nuance the same body of research adds. The SBA’s Office of Advocacy and other analyses of the BTOS series note that large enterprises used AI at 1.8 times the rate of small firms in early 2024, but by mid-2025 small businesses were adopting AI at a faster rate while large-firm adoption plateaued — a reversal of how new technology usually spreads. So the May 2026 reading isn’t a permanent verdict that big companies win. It’s a snapshot of a race where the smallest firms had been closing in, and the latest quarter shows the bigger firms putting a step back on. The lead is real, but it’s not structural.

Why do firms with 20+ employees show up as the “biggest users”? Mostly because they have someone whose job includes figuring this out — an ops manager, a marketing lead, an IT contact — plus the slack to run a pilot that might not pan out. A four-person shop doesn’t have a spare person to evaluate tools, and the owner is already wearing five hats. That’s the real divide the Census data is capturing: not appetite or even cost, but bandwidth and ownership. The good news is that both of those are fixable without a budget, and fixing them is exactly how a small operator closes the gap.

Here’s a practical, low-cost way to think about it. First, assign an owner — even if it’s you, for two hours a week, on the calendar. The Census data’s real lesson is that AI adoption follows whoever is accountable for it; in a tiny firm, that has to be a named person and a recurring block, or it never happens. Second, pick one business function, not the whole business. BTOS measures “use in at least one business function” for a reason — adoption starts narrow. Choose the function that’s both painful and repetitive: customer replies, scheduling, invoicing follow-up, social posts, quote drafting. Third, time-box tool selection to two weeks. The single biggest reason small firms stall isn’t that AI doesn’t work; it’s analysis paralysis over which tool. Set a deadline, pick the one that’s already inside software you pay for (your accounting app, your email, your point-of-sale), and start. Fourth, measure one number for 30 days — hours returned, response time, quotes sent — so you can tell whether to expand or kill it.

You don’t need an enterprise budget to act like the firms pulling ahead. You need their two advantages — a named owner and a narrow first use case — both of which a one-person business can manufacture in an afternoon. The sector data backs this up: the leaders (Information, Finance) aren’t ahead because they bought more expensive AI; they’re ahead because their work is information-heavy and they pointed AI at it early. Most small businesses have at least one information-heavy, repetitive task hiding in plain sight.

If you want help compressing that from “someday” to “this month,” LevelUpLabs.co is a membership built for exactly this kind of operator — entrepreneurs who don’t have a dedicated AI person and need a shortcut. Inside you’ll find prompt libraries you can paste in and run, short video training that skips the theory, ready-to-use checklists for picking one workflow and proving it out, and partner discounts on the tools small firms are already adopting. It’s a way to borrow the “named owner and a plan” advantage that the bigger firms in the Census data have by default.

The Census numbers are a leading indicator, not a scoreboard. Bigger firms are ahead this quarter, but small firms were closing the gap a year ago and can again — adoption tracks ownership and focus, not company size. Name an owner, pick one function, set a two-week deadline, and measure the result. That’s the whole gap, and it’s closeable this month.

Sources:

• Large Firms With at Least 20 Employees Biggest AI Users — U.S. Census Bureau
• Census Data Shows Smallest Firms Leading Small-Business AI Adoption — Self Employed
• Research Spotlight — AI In Business: Small Firms Closing In (SBA Office of Advocacy)
• Census: AI Adoption Accelerating Across U.S. Businesses, but Use Cases Remain Narrow — Modern Distribution Management

On May 21, 2026, California Governor Gavin Newsom signed a first-of-its-kind executive order directing the state to prepare workers, small businesses, and communities for the economic disruption that artificial intelligence is expected to bring to the workforce. It does not create any new laws, mandate any worker protections, or restrict the use of AI in any business operation. What it does instead is launch a 180-day, data-driven process to figure out what those policies should look like — and to small business owners watching from outside California, that process is more important than any single rule that will eventually come out of it.

The order mobilizes state agencies, labor experts, economists, universities, and industry leaders to develop new policies, gather data, and identify early warning signs of workforce disruption. Specific areas the state has been directed to study include severance standards, employment insurance and transition support, worker-ownership models, universal-basic-capital concepts, expanded workforce training, and stronger tracking of hiring and payroll trends. State agencies have 180 days to come back with concrete recommendations — including possible updates to California’s existing Worker Notification Law, which requires advance warning before mass layoffs.

Concretely, four implementation tools start work immediately. The Employment Development Department is building a public dashboard that will track AI’s impact by sector. The state is producing an “AI playbook” intended to modernize the state’s job training programs. A single online portal will consolidate the state services workers and small businesses use most often. And — this is the line most small business owners should highlight — new business-feedback inputs are being wired directly into California’s monthly jobs report, so the state can react to AI-driven labor shifts in close to real time rather than catching up after a lagging quarter of unemployment claims.

Why does this matter if you don’t run a business in California? Three reasons.

One: California has historically been the leading-indicator state on workforce and consumer regulation. From minimum wage benchmarks to CCPA-style privacy law, what California writes tends to define the floor that other large states (New York, Illinois, Colorado, Washington) write toward inside 18 to 36 months. Colorado SB 26-189 just rewrote its landmark AI law and is expected to take effect January 1, 2027. The federal AI for Main Street Act is now law, the GAO published its first AI-and-small-business-contracting report on May 4, 2026 (GAO-26-107828), and the House Committee on Small Business unanimously advanced nine bipartisan bills in the week of May 18, 2026 — two of them AI-specific. The regulatory environment around AI and labor is converging on something, and Newsom’s executive order is one of the clearest signals yet of what it is converging toward: data-driven, sector-specific, transition-focused, with an explicit small-business carve-out.

Two: the EDD AI-impact dashboard, when it ships, will be the first public dataset of its kind. Most small business owners are flying blind on the macro question of which sectors AI is actually displacing labor in, on what timeline. If you employ even five people, a sector-by-sector view of AI labor impact is more useful than another LinkedIn think piece. Bookmark it the day it goes live.

**Three: this is the first executive-branch document to explicitly name small businesses alongside workers as a population the state will protect against AI disruption.** That framing matters for how grants, training credits, SBA-style loan guarantees, and procurement preferences will be designed over the next 18 months — both inside California and at the federal level. The signal to operators is to start thinking of AI policy as something with potential upside for small businesses (training subsidies, financeable AI tooling under H.R. 915, eventual procurement set-asides) rather than purely as compliance overhead.

If you want a place that translates policy moves like this into the workflows, prompts, and partner discounts that small business operators can actually use, take a look at LevelUpLabs.co. It’s a membership built for entrepreneurs and small business owners who want to put AI to work without first earning an MBA in regulatory affairs — packed with prompt libraries, video training, ready-to-use operating checklists, and exclusive partner discounts on the AI tool stack that’s already on the procurement standards conversation in Sacramento, Washington, and your state capital next.

Here is the four-step operator response. Step one — read the order itself, not the headlines. The actual EO language is short, and the sections on small business inputs into the monthly jobs report and the EDD dashboard tell you where the data will live. Step two — write down your own internal AI/labor policy in one page. Which roles in your company are you augmenting with AI? Which are you backfilling? What does your version of “advance notice” look like for any role you genuinely intend to retire? Doing this now, voluntarily, is a tiny lift that puts you ahead of any rule that lands in 2027. Step three — keep your training and reskilling budget visible as a line item. Several of the federal bills moving through Congress (H.R. 915 Small Business Technological Act, the Small Business AI Training Act, the AI for Main Street Act grant program) point at training as a financeable and grant-eligible category. If your books show real training spend, you’ll qualify when those programs go live. Step four — pick one human-judgment role in your company and protect it explicitly in writing. The throughline of every credible policy document released in the last 60 days is the same: AI augments, humans approve. The companies that document where the human approval lives are the ones that will be on the right side of the next compliance cycle and the right side of the customer trust cycle.

The closing takeaway: California just put a 180-day clock on the question every small business owner has been quietly avoiding — which jobs in your company will look different in 24 months, and what do you owe the people doing them. You don’t have to wait for the state to answer it. The operators who answer it themselves, in writing, this quarter, will look like leaders when the rest of the country starts asking.

Sources:

• Office of the Governor of California — Governor Newsom signs first-of-its-kind executive order to prepare workers and businesses for potential AI disruption — https://www.gov.ca.gov/2026/05/21/governor-newsom-signs-first-of-its-kind-executive-order-to-prepare-workers-and-businesses-for-potential-ai-disruption/
• NPR — CA Gov. Gavin Newsom signed an executive order to protect workers from AI — https://www.npr.org/2026/05/22/g-s1-123671/ca-gov-gavin-newsom-signed-an-executive-order-to-protect-workers-from-ai
• CBS Sacramento — Newsom executive order directs California to prepare for AI job disruption — https://www.cbsnews.com/sacramento/news/gavin-newsom-california-ai-job-disruption-executive-order/
• Better Markets — CA Gov Newsom’s AI Workforce Executive Order’s Data-Driven Framework Focused on Jobs and Small Businesses is a Good Start — https://bettermarkets.org/newsroom/ca-gov-newsoms-ai-workforce-executive-orders-data-driven-framework-focused-on-jobs-and-small-businesses-is-a-good-start/
• Consumer Finance Monitor — Colorado rewrites its landmark AI law: Unpacking SB 26-189 — https://www.consumerfinancemonitor.com/2026/05/12/colorado-rewrites-its-landmark-ai-law-unpacking-sb-26-189-and-what-it-means-for-businesses/
• U.S. GAO — Artificial Intelligence: Uses and Risks for Small Business Contracting and Innovation Research (GAO-26-107828) — https://www.gao.gov/products/gao-26-107828

If you have ever tried to hire one person at a small business, you already know the math. According to data cited in HeroHire’s May 17, 2026 launch, the average small business spends 40+ hours on a single hire — drafting the role, screening résumés, scheduling, interviewing, and onboarding. If you give up and use a recruiter, the typical placement fee runs 15% to 30% of the new hire’s first-year salary. On a $70,000 role, that’s $10,500 to $21,000 — gone before the new hire writes a line of code, takes a call, or pours a cup of coffee.

That is the hiring tax HeroHire — which announced its general availability on May 17, 2026 — is built to attack. The company describes the product not as a job board, applicant tracking system, or dashboard but as an “autonomous AI agent” pointed at small and mid-size companies. The operator talks to it through a voice interface, describes the role and the must-haves, and HeroHire sources candidates across 800+ million profiles drawn from 80+ data sources. The agent itself handles outreach, scheduling, and screening before a human ever opens a tab.

For owners, this is the latest entry in what has now become a recognizable May 2026 pattern. Anthropic’s Claude for Small Business (May 13) bundled HR among 15 agentic workflows. Intuit’s QuickBooks Workforce expansion (May 6) absorbed payroll and benefits handoffs into an agent. Microsoft Agent 365 went GA on May 1. Square Managerbot opened beta scheduling for restaurants in April. LinkedIn’s Hiring Pro added a plain-language AI agent in Applicant Evaluation on May 12. Every category of the SMB back office is being absorbed into an agentic layer, and hiring is among the last to fall.

It’s worth asking why hiring resisted automation longer than payroll, scheduling, or invoicing. The honest answer is that the workflow was never the bottleneck — judgment was. Whether to call back a candidate who has the wrong degree but the right portfolio, whether to push past the awkward third interview, whether to hire for the role you have or the role you’ll need in six months — none of that is a list-sorting problem. What changed in 2026 is that the surrounding work — sourcing, scheduling, first-pass screening, calendaring follow-ups — is now genuinely automatable to a level where a small business owner can spend most of the saved 40 hours on the judgment part instead.

So how should an owner evaluate this — not just HeroHire specifically, but the broader autonomous-recruiter category? A clean 4-step framework:

(1) Price the agent against the labor it actually replaces. If you would have used a recruiter, the comparison is platform cost vs. the 15%–30% placement fee on the role’s first-year comp. If you would have done it in-house, the comparison is platform cost vs. 40 owner-hours valued honestly (use your fully-loaded billable rate, not minimum wage). Two roles a year of recruiter savings will pay for any of these tools several times over.

(2) Demand the human-approval gate. No autonomous agent should send an offer letter, post a salary band, or decline a candidate without you clicking approve, especially for the first 90 days. HeroHire’s own framing emphasizes the owner staying in the loop on must-haves; honor that on your side by reading every proposed shortlist and outreach message for the first ten roles. This is also the period you’ll catch the agent’s blind spots — schools it overweights, keywords it misreads, biases that show up in its candidate set.

(3) Audit for protected-class risk and explainability. Federal EEOC guidance and a growing patchwork of state AI hiring laws (New York City Local Law 144 was the early model; Colorado, Illinois, and California have follow-ons in flight) require employers using automated decision tools in hiring to document them. Before going live, get from the vendor: (a) what the agent uses to score candidates, (b) how it handles missing data, (c) whether it audits for adverse impact, and (d) what your recordkeeping obligations are. A vendor that can’t answer those four questions is not a vendor — it’s a lawsuit.

(4) Re-baseline the hiring calendar. The most common mistake an owner makes after automating a workflow is leaving the old cadence in place. If your prior baseline was “open a role, expect 60 days to first hire,” your new baseline with an autonomous recruiter should be 14 to 21 days end-to-end. Plan the onboarding, the seat, the laptop, and the first-week schedule on the new timeline, not the old one. The cost of running ahead of the agent is empty seats; the cost of running behind it is the new hire showing up with nothing to do.

If you want the operator-side playbooks for the rest of this — how to write a job description an AI agent can actually use, how to set the must-have list so the shortlist doesn’t come back full of false positives, how to plug a HeroHire-style agent into a Claude for Small Business or QuickBooks Workforce setup so onboarding doesn’t break — LevelUpLabs.co is the membership for it. Prompt libraries, video walkthroughs, hiring checklists, and exclusive partner discounts on the SMB AI stack, all built for small business owners who want to put these tools to work this week instead of reading another think-piece about them next week.

The closing takeaway. The HeroHire launch isn’t a story about one product. It’s a marker. Hiring just joined finance, scheduling, marketing, and ops as a workflow with a small-business-priced autonomous agent option. Whether you adopt this specific tool or wait for the next one, the operator move is the same: take the 40 hours you currently lose per hire, the $10–21K per-hire recruiter fee you currently swallow, and decide which two hires this year you’re going to test an autonomous recruiter on. Then run the 4-step framework above. The owners who do this in 2026 will hire faster and cheaper than the ones who don’t, and the gap will compound.

Sources:

• GlobeNewswire — HeroHire Launches Autonomous AI Recruiter to Fix Hiring for the 99% (May 17, 2026) — https://www.globenewswire.com/news-release/2026/05/17/3296256/0/en/HeroHire-Launches-Autonomous-AI-Recruiter-to-Fix-Hiring-for-the-99.html
• Paychex — AI Tools for Recruiting: 2026 Small Business Guide — https://www.paychex.com/articles/human-resource/ai-tools-for-recruiting
• Entrepreneur — How LinkedIn Hiring Pro’s AI is Changing Small Business Hiring in 2026 — https://www.entrepreneur.com/business-news/how-ai-is-changing-small-business-hiring-in-2026/502054
• Anthropic — Introducing Claude for Small Business — https://www.anthropic.com/news/claude-for-small-business
• Intuit — QuickBooks 2026 AI Impact Report — referenced across May 13, 2026 coverage
• Federal Reserve — Monitoring AI Adoption in the U.S. Economy — https://www.federalreserve.gov/econres/notes/feds-notes/monitoring-ai-adoption-in-the-u-s-economy-20260403.html

There are two ways to read the headline number from Goldman Sachs’s latest small business survey, and the difference between them is the whole story.

The optimistic read: 76% of small businesses now say they use artificial intelligence. The sobering read: only 14% say they have fully integrated AI into their core operations. A new survey of 1,256 participants in the Goldman Sachs 10,000 Small Businesses program — conducted by Babson College and David Binder Research — lands squarely on that gap. Most small businesses have AI in the building. Very few have it running the business.

Adoption is not integration

It’s worth being precise about what “76%” actually means, because it’s easy to over-celebrate. Using AI, for most owners, looks like an occasional ChatGPT session, a drafted email, a logo idea, a research question answered faster. That’s real, and the survey backs up the value: among AI users, 93% report a positive impact on their business, and 84% point to increased efficiency and productivity as the main benefit.

But “I use AI sometimes” and “AI is integrated into how my company operates” are different maturity stages. The first is a tool you reach for. The second is a workflow you depend on — AI embedded in how invoices get chased, how customer questions get triaged, how the books get closed, how leads get followed up. Only one in seven small businesses has reached that second stage. The other 62% who use AI are, in effect, idling in first gear.

Why the gap exists

The survey doesn’t leave the reasons to guesswork. Among small businesses using AI, the most common obstacles were concerns about data privacy and security (50%), a lack of technical expertise (49%), and difficulty choosing the right AI tools (48%).

Notice what’s not on that list: cost, or doubt that AI works. Owners aren’t stuck because they think AI is useless or unaffordable. They’re stuck on confidence and clarity — is this safe with my customer data, do I know enough to set it up properly, and which of the forty tools claiming to do this should I actually pick? Those are solvable problems, but they’re not solved by buying another subscription. They’re solved by knowledge.

That’s why the most actionable finding in the survey is the one about demand: more than 70% of respondents said they would benefit from more training and implementation resources. Small business owners aren’t asking for more AI products. They’re asking for help crossing the gap from dabbling to operating. Eighty-five percent said they support the AI for Main Street Act, the federal effort to route AI training and outreach through the SBA and Small Business Development Centers — a clear signal of where owners think the missing piece is.

What an operator should do with this

The practical takeaway is not “adopt more AI.” It’s “integrate the AI you’ve already adopted.” A few concrete moves close the gap the survey describes.

Pick one core workflow — not a side experiment, but something that happens every week and costs you real hours. Invoicing follow-up, customer inquiry triage, and content scheduling are common starting points. Move that one workflow from “I sometimes use AI for this” to “AI runs the first draft of this every time, and I approve it.” One fully integrated workflow beats ten occasional ones.

Address the privacy concern head-on rather than letting it quietly stall you. Read what your AI vendor does with your inputs, keep customer financial and personal data out of consumer-grade tools unless the terms explicitly permit it, and write a one-page internal rule for your team on what can and can’t be pasted into an AI tool. That document is cheap and it removes the single most cited barrier.

Treat tool selection as a time-boxed decision, not an open research project. Difficulty choosing is the third barrier for a reason — owners research forever and integrate nothing. Give yourself two weeks, trial two options for your one chosen workflow, and commit. The cost of a slightly suboptimal tool is far lower than the cost of another quarter spent deciding.

Putting it into practice

The survey’s clearest message is that the bottleneck is training, not technology — owners want help turning AI from a gadget into infrastructure. That’s exactly the gap LevelUpLabs.co is designed to close: a membership for entrepreneurs who want working AI income systems, with step-by-step video training, prompt libraries, implementation checklists, and partner discounts on vetted tools. It’s built for the 62% who use AI but haven’t yet wired it in — a structured path from “I tried it” to “it runs.”

Bottom line

The Goldman Sachs survey shows small business AI adoption is nearly universal and AI integration is still rare. The businesses that pull ahead over the next year won’t be the ones that use AI — almost everyone does. They’ll be the ones that picked a core workflow, handled the privacy and tool-choice questions like adults, and actually moved into the 14%.

Sources:

• Goldman Sachs — Survey: Small Businesses Embrace AI — But Need Training and Support to Fully Harness It (2026)
• Goldman Sachs 10,000 Small Businesses — Technology & Innovation Survey (2026)
• Fortune — Fewer than 1 in 5 small businesses are good at actually integrating AI

If your small business sells to the federal government — or has ever applied for an SBIR or STTR research grant — a report the Government Accountability Office released on May 4, 2026 deserves a few minutes of your attention. It’s not flashy. It won’t trend. But it describes a change happening to the systems that decide whether your bid wins and whether your grant proposal gets funded.

The report, GAO-26-107828, examines how the Small Business Administration, agency small-business contracting offices, and federal innovation research programs might use AI. The short version: AI is poised to enter the review pipeline on the government’s side of the table — and that has direct, practical consequences for the businesses on the other side.

What the report actually says

GAO lays out where AI could plausibly be used. For offices that oversee small-business contracting, AI tools could help with market research — identifying broader pools of small-business contractors, summarizing industry trends — and could assist in compiling acquisition forecasts and building internal training. For the SBIR and STTR innovation programs, the potential uses are even more pointed: AI could help match proposal reviewers to technical submissions, check proposals for compliance, assist with due diligence, detect potential fraud indicators, and refine the feedback sent to applicants who don’t get funded.

GAO is careful about the risks, too. It flags inaccurate outputs, data security concerns, and the possibility of biased outcomes — the same cautions any responsible operator would apply before letting software make consequential calls.

There’s also a pointed accountability finding. GAO found the SBA had not consistently complied with federal requirements to publicly report its AI use-case inventory. The SBA didn’t publish its first such inventory until March 2026, even though reporting requirements have existed in various forms since 2020. GAO recommended the SBA establish policies to fix that. In plain terms: the agency is moving toward AI faster than it’s disclosing how it uses it.

Why this matters to a small business operator

Two practical implications stand out.

First, your proposals and bids are increasingly being read by software before — or alongside — a human. If an AI compliance check screens your SBIR proposal, then formatting, completeness, and following the solicitation’s exact structure stop being clerical details and become pass/fail gates. A proposal that a sympathetic human reviewer would have read generously may simply get flagged by an automated check. The defensive move is boring and effective: follow the solicitation instructions literally, answer every required element explicitly, and don’t bury responsive content in clever prose.

Second, the transparency gap is itself the story. When an agency adopts AI faster than it documents that adoption, applicants lose the ability to know how decisions are being influenced. You can’t appeal a process you can’t see. That’s not a reason for alarm — GAO’s recommendation exists precisely to close the gap — but it is a reason to read agency AI use-case inventories when they’re published, and to ask procurement officers directly how AI factors into evaluation when it’s appropriate to ask.

There’s an upside, too. The same report notes AI could help agencies identify broader pools of small-business contractors. If you’ve struggled to get on a buyer’s radar, better market-research tooling on the government side could surface your business in searches it previously missed. Make that easier: keep your SAM.gov registration current, your capability statement sharp, and your NAICS codes accurate, because that’s the data those tools will read.

The same lesson applies inside your own walls

The most useful thing in the GAO report for a non-contractor is the framing. GAO didn’t say “use AI” or “don’t” — it said map the specific tasks where AI helps, name the risks honestly, and document what you’re doing. That’s a sound model for adopting AI in any small business: pick concrete tasks, acknowledge the failure modes, and keep a record of how the tool is used so you can audit it later.

Turning that principle into a working routine is what LevelUpLabs.co is built to help with — a membership where small-business owners get AI strategy frameworks, prompt libraries, video training, and ready-to-use checklists, plus partner discounts on the tools worth deploying. It’s a faster path from “I should probably do something about AI” to a documented, repeatable process.

The takeaway

You don’t need to do anything dramatic this week. But if government revenue is part of your business, treat this report as a heads-up: the evaluation process is getting automated components, compliance precision now carries more weight, and the agencies are moving faster than their own disclosure. Tighten your proposal discipline, keep your federal registrations clean, and watch for the SBA’s AI use-case inventory. The small businesses that adapt to how decisions get made will keep winning work — quietly, while everyone else is still reading the headlines about chatbots.

Sources:

• GAO — Artificial Intelligence: Uses and Risks for Small Business Contracting and Innovation Research (GAO-26-107828)
• AI and Small Business Contracting: GAO Identifies Promise, Risk, and a Transparency Gap at SBA — Fed Contract Pros
• GAO Says SBA Did Not Consistently Report AI Use Cases — ExecutiveGov

Most AI news that lands on a small business owner’s desk is a product launch. This one is different — it’s policy, and it has the potential to change something genuinely practical: how you pay for the AI tools you’re already being told you need.

During the week of May 18, 2026, the House Committee on Small Business held a markup and unanimously reported nine bipartisan bills favorably to the House. Two of them speak directly to artificial intelligence, and one of those two could matter to your budget far more than the next chatbot release.

The bill that affects your wallet

The headline for operators is H.R. 915, the Small Business Technological Act, which the committee advanced by a 23-0 vote. The bill would clarify that the SBA’s flagship 7(a) loan program can be used by small businesses to access modern business software, cloud-computing services, AI tools, and other technologies that power day-to-day operations.

That clarification sounds dry. It isn’t. The 7(a) program is the most widely used SBA-backed lending channel in the country, and for years it has been understood primarily as a vehicle for equipment, real estate, working capital, and acquisitions. Explicitly naming software, cloud services, and AI tools as eligible uses signals — to lenders and to borrowers — that a technology stack is now financeable infrastructure, not a discretionary expense you scrape together from cash flow.

For a small business weighing a meaningful AI investment — a customer-service agent, an analytics layer, an automation platform, the training to actually use them — the question has often been “can the business absorb this right now?” A clearer financing path reframes that question. It doesn’t make the tools free, and it doesn’t make every purchase wise, but it removes the assumption that AI adoption has to be self-funded out of a tight month.

The bill about how government uses AI

The second AI-focused bill, H.R. 8881, the SBA Artificial Intelligence Utilization Act of 2026, comes from Reps. Brad Finstad (R-Minn.) and George Latimer (D-N.Y.) and also passed with unanimous committee approval. It would require the SBA Administrator to deliver a report to Congress — within 90 days of enactment, then annually — on the agency’s own use of AI and machine learning, including the benefits, the risks, and related concerns. The bill leans on existing definitions from the National Defense Authorization Act for Fiscal Year 2021 and emphasizes that human judgment must remain central to decision-making.

This one won’t change your operations directly. But it’s worth noticing what it signals: the federal agency that backs small business lending is being pushed to be transparent and disciplined about its own AI use. Owners watching how the SBA processes loans, reviews applications, and flags risk should expect more visibility into where automation is — and isn’t — involved in decisions that affect them.

How an operator should read all of this

A few practical points.

First, don’t wait on legislation to make a decision you’ve already justified. A bill that’s been reported favorably out of committee still has to clear the full House, the Senate, and the President. If an AI tool already pays for itself in hours saved or revenue gained, the financial case stands on its own today. Treat the legislation as a possible tailwind, not a reason to delay.

Second, talk to your lender now, not later. If financing is the real obstacle to an AI investment, ask your SBA-preferred lender how they currently treat software and technology costs inside a 7(a) package. Many already accommodate it within working capital. Knowing your lender’s posture today means you can move quickly if the rules get more explicit.

Second-and-a-half, keep the human-judgment principle for yourself. Congress wrote “human judgment remains integral” into a bill about a federal agency. It’s equally good policy for a five-person company. Whatever you automate, keep a person accountable for the decisions that touch money, customers, and compliance.

The hard part was never deciding that AI matters — it’s turning that decision into a system that actually earns its cost. That’s the gap LevelUpLabs.co is built to close: a membership for entrepreneurs who want working AI income systems instead of more headlines, with prompt libraries, video training, practical checklists, and partner discounts on the tools you’d be evaluating anyway. Financing the stack is one question; getting a return on it is the one that actually pays the loan back.

Bottom line

The committee markup is a reminder that AI is no longer just a product story for small business — it’s becoming a policy story too. The most useful piece for operators is the move to make AI tools clearly financeable through the 7(a) program. Watch it advance, talk to your lender, and keep building the return case in the meantime. Legislation can lower the cost of adopting AI. It can’t do the adopting for you.

Sources:

• House Committee on Small Business — House Committee on Small Business Unanimously Reports Nine Bipartisan Bills Favorably to the House (May 2026)
• FedScoop — House lawmakers seek more AI transparency from the SBA
• Congress.gov — H.R.8881, SBA Artificial Intelligence Utilization Act of 2026
• BizSugar — House Small Business Committee Advances Nine Key Bills to Support Local Enterprises

Most small business owners think of ChatGPT as a chat box — somewhere you ask a question and get an answer. With OpenAI’s Workspace Agents, that mental model is now out of date, and the change carries a cost most owners haven’t budgeted for. Workspace Agents are shared, always-on agents — successors to the older custom GPTs — that plug into Slack, Salesforce, and more than 60 other business apps. They run in the cloud, so they keep working when you’re not. And as of May 6, 2026, the free preview period ended: Workspace Agents now run on credit-based pricing. ChatGPT didn’t just add a feature. It became an automation platform, and it turned on a meter.

Here’s what that actually means in practice. A Workspace Agent isn’t a conversation — it’s a standing worker. You build it once, share it across your team, and it handles repeatable jobs: preparing reports, drafting and sending messages, pulling data across connected apps, executing multi-step tasks on a schedule. It’s available across ChatGPT Business, Enterprise, and Edu plans. The capability is genuinely useful for a small team — it’s the closest thing to hiring an operations assistant without payroll. But because agents run continuously rather than only when you type, their cost scales with how much work they do, not with a flat seat fee. That’s the part owners need to plan for.

This isn’t an OpenAI quirk. It’s the direction the whole category is moving. Through 2026, nearly every tool a small business already pays for — from CRMs to accounting software to help desks — has shipped an agentic layer, and the metered, usage-based pricing model is becoming standard. The convenience is real. So is the risk of waking up to a bill driven by an agent that ran far more often than you expected. Here’s a four-step framework for adopting agents without losing control of the budget.

1. Start with one painful, well-defined workflow. Don’t deploy agents broadly because the capability exists. Pick a single recurring task that genuinely drains time — the Monday report you assemble by hand, the routine customer follow-ups, the data you copy between two apps every week. A narrow, well-scoped agent is easy to measure and easy to cap. A vague one is how usage costs spiral.

2. Price the agent against the labor it replaces. Before you scale anything, run it for two to four weeks and compare two numbers: the credit cost the agent consumed, and the hours of human time it gave back. If an agent costs you the equivalent of a few hours of wages but returns a full day, that’s a clear win. If you can’t articulate the trade, you’re not ready to expand it. Treat every agent like a small hire that has to justify its seat.

3. Set explicit limits and approval gates. Decide in advance how often an agent should run and what it’s allowed to do on its own versus what needs your sign-off. Anything touching money, customer commitments, or public-facing communication should propose an action and wait for a human to approve it — not execute silently. Review every agent’s output for the first 30 days so you learn its failure patterns before you trust it.

4. Re-budget AI as a variable operating cost. The old model was a fixed monthly software subscription. The new model is closer to a utility bill — it moves with usage. Build a line item for AI agent spend, check it monthly the way you’d check a payroll or hosting line, and set an internal ceiling that triggers a review if it’s crossed.

If you want help deciding which workflows are worth automating — and how to do it without overspending — that’s the gap LevelUpLabs.co is designed to close. The membership gives small business owners practical AI strategies, a library of prompts and agent templates, video training, and operator checklists, plus partner discounts on the tools themselves — so you can put agents to work deliberately instead of paying tuition through trial and error.

The takeaway: Workspace Agents are a real upgrade — a way for a small team to get standing operational help without adding headcount. But the shift from free preview to a metered model means AI has quietly moved from a fixed software cost to a variable one. Owners who treat agents like employees — scoped, measured, supervised, and budgeted — will get the leverage. Owners who treat them like a free toy will get a surprising invoice. Pick one workflow this week, run the math, and decide from there.

Sources:

• Introducing workspace agents in ChatGPT — OpenAI
• OpenAI unveils Workspace Agents, a successor to custom GPTs for enterprises — VentureBeat
• OpenAI Replaces Custom GPTs With Workspace Agents Built for Team Workflows — Reworked
• ChatGPT Business — Release Notes — OpenAI Help Center