Which AI Consulting Company Should You Choose?

Six alternatives are realistic: a boutique implementation firm, an embedded build partner, an independent or fractional lead, the IT partner you already pay, the software you already run, and free or low-cost public programs. For an operations-heavy business under 200 people, the answer is almost always one of the middle three. One of those three is our own model, so weigh what follows accordingly. Here's the number that explains why. Median monthly spending on AI services was "approximately $80 per month in 2022 before declining to roughly $30 per month by 2025," measured from actual payments across 4.6 million small businesses (JPMorganChase Institute, Understanding the use of AI among small businesses, April 2026). Now set that against what a large firm publishes for the same category of work. On the UK's G-Cloud 14 framework, suppliers have to post a price. KPMG LLP lists its Data Science and AI Capability Adoption Partner service there at £325 to £2,325 a person a day (Crown Commercial Service, G-Cloud 14 service listing, retrieved 2026-08-17). One person. One day. That gap isn't evidence that big firms are bad at this. It's evidence they're built for someone else, and the useful move is to stop shopping for a smaller version of them.

Key Takeaways

• Median monthly AI spend at a US small business fell to roughly $30 by 2025 — real payments, not a survey (JPMorganChase Institute, 2026).
• On the one framework where these firms publish prices, top-of-range day rates hit £2,240 (Accenture), £2,750 (PwC) and £2,325 a person (KPMG).
• Don't look for a cheaper firm. Unbundle the engagement: the diagnostic, the roadmap, the build and the accountability can each be bought separately.
• Deloitte's own 2026 survey found just 25% of organizations had moved 40% or more of their AI pilots into production.
• Mid-market firms reached implementation in about 90 days, against nine months or more at large enterprises (MIT Project NANDA working paper, 2025).

Table of Contents

• What Are the Alternatives to a Big AI Consulting Firm?
• What You're Actually Buying When You Hire a Big Firm
• The Only Day Rates These Firms Publish Anywhere
• Where the Big-Firm Model Breaks at 10 to 200 People
• Which Alternative Replaces Which Part of the Bundle?
• Is Hiring In-House Cheaper Than Any of Them?
• When Is a Big Firm Actually the Right Answer?
• How Do You Find and Vet the Alternatives?
• When Should You Walk Away?
• Your Next Step
• Frequently Asked Questions
• Sources
• About Avolis Research Group
• Continue Learning

What Are the Alternatives to a Big AI Consulting Firm?

Six routes are genuinely available to a business your size. They differ in price, and just as much in what you have to supply yourself. Start with the shape of the market. IBISWorld estimates 1,210,001 US management consulting businesses against 2,565,423 people employed, and puts the average business at 2.1 employees (IBISWorld, Management Consulting in the US, August 2026). That's a commercial estimate rather than a government count, and the business tally includes sole proprietors. The brand names are still the exception, not the market. Here's the field.

Two notes before the detail. Nothing on that list is a downgrade. Most of it is what businesses of this size actually use, and the last two rows are frequently free. The routes also combine, which is the part no vendor will mention. A diagnostic from a public program, a build from a boutique firm, and upkeep from your existing IT partner is a completely sane sequence.

Want the wider version of this decision? Our guide to which AI consulting company you should choose covers how to weigh firm type against your own readiness.

What You're Actually Buying When You Hire a Big Firm

A large engagement is four products sold as one. You're buying a diagnostic, a roadmap, build capacity, and accountability — and the price is set by the bundle, not by the part you needed. Deloitte surveyed 3,235 business and IT leaders across 24 countries in its 2026 study. Only 30% of organizations were redesigning key processes around AI, while 37% were "using AI at surface level with minimal business process changes" (Deloitte, The State of AI in the Enterprise: The Untapped Edge, January 2026). Most buyers are paying for a bundle and using one corner of it.

Our finding: "what's a cheaper alternative to a big firm?" is the wrong question, and it's why so many buyers end up disappointed by a smaller firm too. You're not replacing a vendor. You're replacing four things that were priced together — the diagnostic that finds the work, the roadmap that sequences it, the hands that build it, and the named person who owns the result in month four. Price each one separately and you stop paying for the parts you weren't going to use. Most operations-heavy businesses need a real diagnostic, a real build, and almost none of the roadmap. The roadmap is the expensive part. It's also the part that goes stale fastest.

Look at what happens to each component when you buy it à la carte. The diagnostic is free from a Small Business Development Center, cheap from a manufacturing extension center, or a fixed fee from an implementation firm. Build capacity is available from a two-person shop. Accountability is a contract term, not a brand. Only the roadmap has no cheap substitute — and for a 40-person operation running one or two workflows through AI, a roadmap is a document you don't need.

That's the whole argument, and it has a limit worth stating. Unbundling works on narrow scope. It stops working when the scope is genuinely enterprise-wide, which we come back to further down.

The Only Day Rates These Firms Publish Anywhere

Large consultancies don't publish US price lists, and the hourly figures that circulate online are almost entirely unsourced. There's one exception. To sell into the UK public sector, suppliers on the G-Cloud 14 framework must post a price on a government domain, and several large firms have published rates for AI work specifically.

What we did, so you can check it: on 17 August 2026 we opened the G-Cloud 14 listings for AI services from Accenture, PwC, KPMG and Deloitte, then recorded the published price string on each page. Accenture (UK) Limited lists both its AI and Advanced Analytics Services and its Secure AI Services at "£95 to £2,240 a unit a day." PricewaterhouseCoopers LLP lists AI Integration and Implementation Services at "£100 to £2,750 a unit a day." KPMG LLP lists its Data Science and AI Capability Adoption Partner at "£325 to £2,325 a person a day." Deloitte LLP publishes a single flat figure of "£450 a unit a day" on two AI listings rather than a range. Four firms, six listings, one date. Anyone can reproduce it in ten minutes.

Read those numbers carefully, because the caveats matter more than the headline. The top of each range is a ceiling for partner and director grades. Nobody should write that Accenture charges £2,240 a day. "A unit a day" is the supplier's own declaration and isn't necessarily one person. Only the KPMG listing says "a person a day," which makes it the cleanest per-head figure of the six. These are UK public-sector framework prices in pounds, ex-VAT. We're not converting them to dollars and we're not presenting them as US commercial rates, because we have no verifiable source for those. What they do establish is a floor of published fact in a category otherwise full of guesses.

Now put the two ends of this article's evidence side by side. Roughly $30 a month is what the median small business actually pays for AI services. Between £325 and £2,325 buys one person for one day at KPMG. Both figures are published and sourced. They also measure different things — software subscriptions on one side, senior human labour on the other — and that's the point rather than a flaw in the comparison. Your budget is anchored to the first number. The proposals you're reading are priced off the second. That distance is why "which big firm should we call?" is rarely the right opening question. For a closer look at that end of the market, see what the Big Four charge for AI tool development. For pricing models across the whole field, start with how much AI consultants charge.

Where the Big-Firm Model Breaks at 10 to 200 People

It breaks on client size, and the firms' own reporting shows it plainly. Accenture serves roughly 9,000 clients, and ended fiscal 2025 with about 779,000 people on $69.67 billion of revenue (Accenture, newsroom fact sheet, retrieved 2026-08-17). Divide the revenue by the client count and the average client is worth about $7.7 million a year. That division is ours, from two figures on one page. In the first quarter of fiscal 2026 alone, the firm recorded 33 clients with quarterly bookings above $100 million (Accenture, First-Quarter Fiscal 2026 Results, December 2025). This is a business organized around nine-figure programs. That's not a knock on the firm. It's arithmetic: a 40-person contractor sits far below the engagement size this machine is built to run.

The capability is real, though, and that deserves saying without hedging. Accenture booked $5.9 billion in new generative-AI business in fiscal 2025. It recognized $2.7 billion of generative and agentic AI revenue in the same year (CIO Dive, September 2025). Against $69.67 billion of total revenue, that's about 3.9% — again our arithmetic, from two published numbers. Even at the loudest AI shop in consulting, that reported line sat under four percent of revenue last year. Its wider AI and data practice is much larger than the generative-and-agentic figure it breaks out.

The second break is the handoff itself. Deloitte's 2026 survey found only 25% of organizations had moved 40% or more of their AI pilots into production. Only 20% were currently achieving revenue growth from AI, against 74% who aspired to (Deloitte, January 2026). Sit with what that is. A firm the category is named after, surveying its own market, reporting that plans mostly don't become systems.

Here's the part that surprised us most. The largest consulting buyer on earth reached the same conclusion and acted on it. In 2025 the US General Services Administration told agencies that the ten highest-paid consulting firms were "set to receive over $65 billion in fees in 2025 and future years," then demanded cost-reduction proposals (Federal News Network, February 2025). GSA's Federal Acquisition Service commissioner, Josh Gruenbaum, said the agency was "unanimously unimpressed." He called the firms' proposals "wholly insufficient, to the point of being insulting" (FedScoop, April 2025). He also told them recommendations should use language "a 15 year old should be able to understand" and avoid "jargon or gobbledygook." By then GSA had canceled 1,700 consulting contracts (Bloomberg, March 2025). One honest qualification: GSA kept buying from the same ten firms, so this was a renegotiation and not a walkout. Even so, if the federal government concluded it was paying for decks, a 60-person business is allowed the same verdict.

One more data point, and it's the most encouraging one here. MIT's own research on the AI value gap found mid-market organizations moved from pilot to implementation in roughly 90 days. Large enterprises took nine months or longer (MIT Project NANDA, The GenAI Divide: State of AI in Business 2025, July 2025). Read that as indicative rather than settled. It's a non-peer-reviewed working paper, and its headline figures have been contested. Still, the direction matches what we see: being small is an advantage here. Fewer approvals, one decision-maker, and a much shorter path from "this is broken" to "this is fixed."

Which Alternative Replaces Which Part of the Bundle?

Each route covers some components and leaves others to you, and knowing which is the whole skill. Census Bureau researchers found that among firms using AI, 57% had it in three or fewer business functions (U.S. Census Bureau, The Microstructure of AI Diffusion, working paper CES-WP-26-25, April 2026). The most common were sales and marketing at 52%, strategy and business development at 45%, and IT at 41%. Real adoption is narrow. Narrow scope is exactly what these routes serve well.

The boutique implementation firm

Covers build capacity, and sometimes a light diagnostic. This is where most businesses that already know their problem should look. The gap it leaves is the diagnostic itself. A boutique firm quotes against the workflow you name, which is fine if you named the right one, and expensive if you named the loudest one rather than the costliest. Buy the diagnostic separately. From anyone.

The embedded build partner

Covers all four components, priced as a diagnostic followed by a retainer. Deloitte's finding that only 30% of organizations are redesigning processes around AI is the case for this model, since redesigning how work happens isn't something anyone hands off in a statement of work and collects later. Disclosure: this is our model, so weigh the recommendation accordingly. We've set out below where it's the wrong call.

The independent or fractional lead

Covers senior expertise, part-time, and nothing else. Fractional Jobs surveyed 1,733 respondents in 2026, 810 of them working fractionally. VP-level-and-above fractional executives reported charging an average of $223 an hour, engineering highest at $229, with 46% billing primarily on a monthly retainer (Fractional Jobs, The Fractional Work Report 2026). That's a marketplace surveying its own network. Its client mix skews to venture-backed startups rather than field-services businesses, so treat $223 as an anchor rather than your quote.

The pool is deeper than most owners assume. MBO Partners counted a record 5.6 million American independent professionals earning more than $100,000 a year, up 19% in a year, in a study of 6,474 US residents (MBO Partners, 2025 State of Independence in America, September 2025). MBO services independent contractors, so that's vendor research about its own market. Heidrick & Struggles surveyed 3,810 full-time independent consultants and interim executives. Small and medium companies "now account for more than four-fifths of demand" for high-end independent talent (Heidrick & Struggles, 2026 Talent Lens Survey: The State of Interim Talent, February 2026). Same caveat: Heidrick owns a marketplace for this talent, and that demand mix is reported by suppliers, not buyers.

A warning about one specific title. If you search for what a fractional Chief AI Officer costs, every figure you'll find is published by someone selling fractional Chief AI Officer services. We looked. There's no survey behind those numbers, so the defensible anchor remains the $223 fractional-executive average above.

What you inherit on this route: buy an independent lead and you've covered exactly one component of the four — senior expertise. The roadmap and the accountability stay on your side of the table, which in practice means somebody at your company runs the project. That's the line item nobody quotes for. Price it in hours of your ops manager's week and put that next to the firm's fee, because that's the honest comparison.

The IT partner you already pay

Covers integration and upkeep, and increasingly wants to cover more. Kaseya surveyed more than 1,000 managed service providers worldwide for its 2026 report. Forty-eight percent ranked AI and automation as their clients' number-one need, while only 13% were generating meaningful revenue from AI services (Kaseya, 2026 State of the MSP Report, April 2026). Kaseya sells software to MSPs, so that's a vendor surveying its own customers. The gap is the point. Your IT partner knows you want this and probably can't build it yet. Ask them directly. If they can't, they're still the right people to keep it running once someone else has.

The software you already run

Covers the build itself, and it's the route researchers actually observe. Federal Reserve Bank of San Francisco researchers examined how small businesses use AI and identified three routes: functionality embedded in software the business already runs, freely available platforms, and specialized industry-specific software (Federal Reserve Bank of San Francisco, Early Findings on Small Business Use of AI, March 2026). Commissioning custom work wasn't one of the three. That finding is descriptive rather than a percentage breakdown, and it rests on the 2024 Small Business Credit Survey, a large but non-probability sample. Before you buy anything, check whether your field-service platform or your accounting package already ships the feature you're about to pay someone to build.

For the diagnostic on its own, public programs are the cheapest real option. Often free. Our guide to the top 5 AI consulting firms for small businesses covers the federal and state programs in detail. If you'd rather see a shortlist by firm than by route, start with the best AI consulting firms for small businesses.

Is Hiring In-House Cheaper Than Any of Them?

Usually not, and the wage data is public. BLS reports a median annual wage of $120,230 for data scientists, $135,980 for software developers, and $140,300 for computer and information research scientists (U.S. Bureau of Labor Statistics, Occupational Employment and Wage Statistics, May 2025, via DOL-sponsored O*NET). One thing worth noticing: there is no BLS occupation code for "machine learning engineer" or "AI engineer." Data scientist and software developer are the standard proxies. That tells you how new this labor market still is.

Wages aren't the cost, though. BLS put private-industry compensation at $46.60 per hour worked in March 2026. Of that, $32.60 was wages and salaries and $14.01 was benefits — a 69.9% / 30.1% split (U.S. Bureau of Labor Statistics, Employer Costs for Employee Compensation — March 2026, June 2026).

Gross the median data-scientist wage up on that ratio and one in-house hire costs roughly $172,000 a year. Do the same for a software developer and it's about $195,000. Both are our arithmetic from two BLS series, not published BLS statistics. The 30.1% benefit share is also an all-occupation average that tends to run lower for high earners, so read these as rough approximations and check the math yourself. Neither figure includes recruiting, equipment, or the management time to keep a technical hire pointed at the right work.

Then there's the hiring risk. You're recruiting for a role you can't interview well. BLS projects data-scientist employment growing much faster than average, from a 2024 base of 245,900 jobs with about 23,400 openings a year (U.S. Bureau of Labor Statistics, Occupational Outlook Handbook, 2024–34 projections, via O*NET). For one or two workflows, buying the expertise beats building a department. Weighing this seriously? We go through it properly in AI development companies versus an in-house team.

When Is a Big Firm Actually the Right Answer?

Sometimes it is, and pretending otherwise would be dishonest. The UK's National Audit Office surveyed government bodies on consultancy use. Eighty-six percent of respondents said consultants provided a valuable contribution, and consultants were judged most valuable "when hired to solve specific problems using expertise that the civil service lacks" (National Audit Office, Government's use of external consultants, November 2025). That's the test, and it's a good one. Specific problem, expertise you genuinely don't have, defined end.

Four situations where a large firm earns its rate:

• You're at the top of the mid-market with a board mandate. A multi-site, multi-country program with executive sponsorship and a real budget is the shape these firms are built for.
• You're in a heavily regulated function. Audit-grade documentation, model risk governance, and regulatory sign-off are things large firms do genuinely well and small shops often can't.
• You need a name on the recommendation. If the decision has to survive a board, a lender, or an acquirer, brand is part of what you're buying. That's a legitimate purchase — just price it knowingly.
• The work is research, not operations. A custom model or novel technical work is a specialist job, not an implementation job.

None of that describes a 60-person HVAC company trying to stop losing quotes in an email inbox. Most of our readers are the second thing.

How Do You Find and Vet the Alternatives?

Start from the workflow, not from a vendor list, and vet on process rather than claims. GTIA assessed the technology channel in July 2026 and found only 20% of organizations have the governance frameworks, formal policies and commercial strategies needed to turn AI adoption into lasting value (GTIA, AI Adoption in the Channel, July 2026). That's qualitative research without a published sample, so treat it as GTIA's characterization rather than a measurement. The practical read is still useful. Most providers are figuring this out alongside you, which makes how they work the only thing worth evaluating.

A five-step method you can run in a fortnight:

1. Name the workflow and the number. One workflow, and what it currently costs you in hours or lost jobs. Everything downstream gets easier.
2. Get a diagnostic before a quote. Free from an SBDC, low-cost from a manufacturing extension center, or fixed-fee from an implementation firm. Whoever prices a build without seeing your operation is pricing a template.
3. Source three routes, not three vendors. One boutique firm, one independent lead, and your existing IT partner. Comparing across routes tells you far more than comparing three firms inside one route.
4. Ask each one what happens in month four. A partner describes iteration, measurement, and who owns what. A vendor describes a handover.
5. Check the software you already pay for. Do this before signing anything. Occasionally it ends the project, which counts as a win.

A vetting rule worth adopting: discount any provider's self-reported client-impact numbers, including ours. A federal audit shows why. The Commerce Department's Inspector General examined how NIST reported the economic impact of its Manufacturing Extension Partnership and found the reporting "inaccurate and unreliable." Forty-eight percent of total sales reported by reviewed centers in FY2022 were unreliable. Return on investment was overstated across multiple years, including by 34% in FY2020. Some centers were even requiring clients to complete impact surveys against federal guidelines (U.S. Department of Commerce Office of Inspector General, NIST Overstated MEP's Economic Impacts, September 2024). That's a government program with an audit function attached. A private firm's impact deck has neither. We recommend MEP centers elsewhere on this page, which makes citing their audit awkward — and is exactly why it belongs here. Ask any provider for a reference who runs the system daily instead.

One more thing to ask about, because it's the strongest evidence we found on what makes AI stick. LSE's Inclusion Initiative, working with Protiviti, found 93% of employees who received AI training use AI in their roles, against 57% of those who didn't. Trained employees reported saving about 11 hours a week versus 5, and 68% had received no AI training in the previous year (Protiviti and LSE, Bridging the Generational AI Gap, October 2025). That's correlational, not causal, and Protiviti sells change-management services, so it has an interest in the finding. Still worth asking every candidate who trains your team, and how. For the longer version of this evaluation, see what makes a good AI consulting partner versus a typical vendor.

When Should You Walk Away?

Walk when the proposal is shaped for a company ten times your size. Enterprise-shaped scope at small-business scale usually means the firm has one template and you're being fitted to it.

Four more signals worth ending a conversation over:

• A price before a diagnostic. Whoever quotes work they haven't seen is guessing, and you'll pay for the guess.
• A guaranteed ROI percentage. Nobody can guarantee a number that depends on your data and your team's adoption. See the audit above for what happens to impact claims nobody checks.
• No named owner after go-live. If the answer to "who fixes this in month four" is a ticketing address, the risk transfers to you at handover.
• Savings that only work by cutting people. If the business case collapses when you keep your crew, it was never a productivity case.

There's also an honest gap in the evidence that nobody advertises. Nobody has measured how often small businesses quit an AI project. The much-quoted 42% abandonment figure is enterprise data. S&P Global Market Intelligence surveyed more than 1,000 respondents in North America and Europe. Of those companies, 42% had abandoned most of their AI initiatives in 2025, up from 17% a year earlier (S&P Global Market Intelligence, via CIO Dive, March 2025). We went looking for the small-business equivalent across Federal Reserve, Census and academic sources. It doesn't exist. Treat the scary numbers as what they are — measurements of companies much larger than yours.

Your Next Step

Don't start with a shortlist. Start by working out which part of the bundle you actually need, because that one answer eliminates four of the six routes above and saves you weeks of vendor calls.

The sequence we'd run in your position is short. Name one workflow, and what it costs you today in hours or lost jobs. Then get a diagnostic — free from an SBDC, cheap from a manufacturing extension center, or fixed-fee from an implementation firm — before anyone quotes you a build. Compare across three routes, not three firms in one route.

Ready to find where AI actually pays off in your operation? Start with an AI readiness assessment — it's the cheapest way to learn which of these six routes you're in. If there's nothing worth automating yet, we'll tell you.

Frequently Asked Questions

What is the cheapest alternative to a big AI consulting firm?

A diagnostic from a Small Business Development Center costs nothing. The software you already run may include the feature you were about to commission. Federal Reserve Bank of San Francisco researchers found small businesses use AI through existing software, free platforms and industry-specific tools. Not custom builds.

How much do independent AI consultants charge compared to big firms?

Closer than you'd expect. Fractional executives at VP level and above averaged $223 an hour in Fractional Jobs' 2026 survey, and KPMG lists AI work at £325 to £2,325 a person a day on the UK's G-Cloud 14 framework. A day of each lands in a similar band. You're buying hours, not a bundle.

Are boutique AI consulting firms as good as the Big Four?

For narrow operational scope, often better. They build rather than advise. Deloitte's own 2026 survey of 3,235 leaders found only 25% had moved 40% or more of their AI pilots into production. For regulated, multi-country programs needing audit-grade governance, a large firm still has the edge.

Should I hire an in-house AI person instead of a consulting firm?

Rarely, for one or two workflows. BLS puts the median data-scientist wage at $120,230. Gross that up by the 30.1% benefit share BLS reports for private industry and the loaded cost nears $172,000 — our arithmetic, not a BLS figure. Buying expertise usually beats building a department.

Can my existing IT provider handle AI implementation?

Ask them, then verify. In Kaseya's 2026 survey of more than 1,000 MSPs, 48% said AI and automation was their clients' top need. Only 13% were earning meaningful revenue from it. Many aren't ready to build yet, though they remain the right partner to keep a finished system running.

Sources

All sources retrieved 2026-08-17.

• Crown Commercial Service, G-Cloud 14 service listing: KPMG LLP, Data Science and AI Capability Adoption Partner — "£325 to £2,325 a person a day." Supplier-published price on a UK government framework; the only one of the six listings stated per person.
• Crown Commercial Service, G-Cloud 14 service listing: Accenture (UK) Limited, AI and Advanced Analytics Services — "£95 to £2,240 a unit a day."
• Crown Commercial Service, G-Cloud 14 service listing: Accenture (UK) Limited, Secure AI Services — "£95 to £2,240 a unit a day," the same range as the listing above.
• Crown Commercial Service, G-Cloud 14 service listing: PricewaterhouseCoopers LLP, AI Integration and Implementation Services — "£100 to £2,750 a unit a day."
• Crown Commercial Service, G-Cloud 14 service listing: Deloitte LLP, AI Scaling & Transformation in Public Sector — "£450 a unit a day," a single figure rather than a range.
• Crown Commercial Service, G-Cloud 14 service listing: Deloitte LLP, Generative AI in Cloud — "£450 a unit a day," matching the listing above. UK public-sector rates throughout, in GBP, ex-VAT; not US commercial rates.
• JPMorganChase Institute, Understanding the use of AI among small businesses, April 2026 — de-identified Chase Business Banking transaction data covering over 4.6 million small businesses, 2019–2025. Behavioural payment data, not a survey; captures direct subscription payments, so it understates AI arriving bundled inside other software.
• Deloitte, The State of AI in the Enterprise: The Untapped Edge, January 2026 — self-reported survey of 3,235 director-to-C-suite leaders, 24 countries, fielded August–September 2025.
• Accenture, newsroom fact sheet — FY25 revenue $69.67bn, ~9,000 clients, and approximately 779,000 people at fiscal-2025 year-end (799,000 as of Q3 FY26). The $7.7m average revenue per client is our own division of two figures on this page.
• Accenture, Accenture Reports First-Quarter Fiscal 2026 Results, 18 December 2025 — 33 clients with quarterly bookings above $100m; advanced AI new bookings $2.2bn.
• CIO Dive, "Accenture completes 'reinvention' as generative AI revenues roll in," 25 September 2025 — FY25 generative and agentic AI revenue of $2.7bn and FY25 generative-AI new bookings of $5.9bn. The 3.9%-of-revenue figure is our own arithmetic.
• Federal News Network, "GSA tells agencies to target top 10 consulting firms for cuts," 27 February 2025 — the "$65 billion in fees" figure and the named list of ten firms.
• FedScoop, "GSA 'unanimously unimpressed' by consulting firms' cost-savings proposals," 17 April 2025 — quotes attributed to Josh Gruenbaum, GSA Federal Acquisition Service commissioner.
• Bloomberg, "Trump official demands 'no gobbledygook' from consultants," 21 March 2025 — the "gobbledygook" instruction and the 1,700 canceled contracts, via a syndicated reproduction of the Bloomberg wire story.
• MIT Project NANDA, The GenAI Divide: State of AI in Business 2025, July 2025 — the 90-day mid-market implementation finding. A non-peer-reviewed working paper (v0.1) mirrored at mlq.ai; no permanent MIT-hosted URL located as of 2026-08-17. MIT's NANDA group: https://www.media.mit.edu/groups/nanda/overview/
• U.S. Census Bureau, The Microstructure of AI Diffusion: Evidence from Firms, Business Functions, and Worker Tasks (CES-WP-26-25), April 2026 — working paper; has not undergone Census review.
• IBISWorld, Management Consulting in the US (NAICS 54161), August 2026 — commercial industry research built on Census and BLS inputs with an undisclosed model. NAICS 54161 is broad, covering HR, marketing and strategy consulting alongside technology work.
• Fractional Jobs, The Fractional Work Report 2026 — 1,733 responses including 810 active fractional workers, recruited from the publisher's own platform network. A marketplace surveying its own market; client mix skews to venture-backed startups.
• MBO Partners, 2025 State of Independence in America, September 2025 — survey of 6,474 US residents including 2,402 independent workers, fielded April 2025, weighted to US demographics. MBO services independent contractors.
• Heidrick & Struggles, 2026 Talent Lens Survey: The State of Interim Talent, 12 February 2026 — online survey of 3,810 full-time independent talent, fielded August 2025. Heidrick owns Business Talent Group, a marketplace for this talent; the demand mix is reported by suppliers, not buyers.
• Kaseya, 2026 State of the MSP Report, 14 April 2026 — survey of more than 1,000 managed service providers worldwide. Kaseya sells software to MSPs.
• GTIA, AI Adoption in the Channel, 22 July 2026 — qualitative research developed from interviews in Q1 2026; no sample size published.
• Federal Reserve Bank of San Francisco, Early Findings on Small Business Use of AI, 23 March 2026 — based on the 2024 Small Business Credit Survey, a large non-probability convenience sample. The three-routes finding is descriptive, not a percentage breakdown.
• U.S. Bureau of Labor Statistics, Occupational Employment and Wage Statistics, May 2025, via DOL-sponsored O*NET OnLine — median annual wages: data scientists (SOC 15-2051) $120,230; software developers (15-1252) $135,980; computer and information research scientists (15-1221) $140,300. O*NET is sponsored by the US Department of Labor and republishes BLS OEWS data. There is no BLS occupation code for machine learning or AI engineer.
• U.S. Bureau of Labor Statistics, Employer Costs for Employee Compensation — March 2026, released 12 June 2026 — $46.60 total per hour worked, $32.60 wages (69.9%), $14.01 benefits (30.1%), all private industry. The ~$172,000 and ~$195,000 fully-loaded estimates are our arithmetic applying this all-occupation ratio to the OEWS medians, not published BLS figures.
• U.S. Bureau of Labor Statistics, Occupational Outlook Handbook, 2024–34 projections, via O*NET — data scientists: 245,900 jobs in 2024, about 23,400 annual openings, growth "much faster than average." Model-based ten-year forecasts, not measurements.
• National Audit Office, Government's use of external consultants, 21 November 2025 — UK central government; sample size not disclosed in the release.
• U.S. Department of Commerce Office of Inspector General, NIST Overstated MEP's Economic Impacts to Congress and Other Stakeholders, 25 September 2024
• Protiviti and the London School of Economics' Inclusion Initiative, Bridging the Generational AI Gap, 28 October 2025 — nearly 3,000 workers and 240 executives surveyed globally; field period not published. Correlational. Protiviti sells AI implementation and change-management services.
• S&P Global Market Intelligence, via CIO Dive, 14 March 2025 — survey of more than 1,000 respondents in North America and Europe. Enterprise data, not small business.

---

About Avolis Research Group

Avolis Research Group is Avolis's in-house research practice, focused on how operations-heavy small and mid-sized businesses actually adopt AI. It synthesizes primary economic research, government survey data, and results from real implementations into practical, vendor-neutral guidance.

More about Avolis and how we work · Get in touch

Continue Learning

• Which AI consulting company should you choose? — the full decision guide this article sits under.
• What the Big Four charge for AI tool development — the pricing end of the big-firm route.
• Best AI consulting firms for small businesses — a shortlist by firm rather than by route.
• Top 5 AI consulting firms for small businesses — how to read published rankings, plus the public programs most owners have never heard of.
• How much do AI consultants charge? — pricing models across the whole field.
• AI development companies vs. an in-house team — the build-versus-buy decision in full.
• What makes a good AI consulting partner vs. a typical vendor — the six-signal evaluation test.
• AI readiness assessment services — how a diagnostic finds the workflow worth fixing.
• Is it worth hiring an AI consultant for a small business? — the honest tradeoffs at small scale.