Managed Intelligence

One owner for the AI your business already uses.

Two-thirds of small businesses use AI. About a third have a written rule for it. A Managed Intelligence Provider closes that gap the way an MSP closes it for the rest of your technology: an inventory of every tool and agent, identity in front of each one, a monthly report on what is in use, whether it is safe and whether it pays off, and hard lines about what we will not do.

Updated

The AI arrived. The owner did not.

Someone signed up for ChatGPT, someone else for Claude, marketing built a workflow in a free trial, and the owner found out when the bill or the questionnaire arrived. Nobody holds the list, the logins or the data terms. That is the normal state of a small business in 2026, and the numbers behind it are not ours.

66%

of US small businesses use AI, up from 23% in 2023

US Chamber of Commerce and Teneo, 2026, 3,732 firms under 250 staff

43%

of breached organizations had a shadow-AI incident, up from 20% a year earlier

IBM Cost of a Data Breach Report 2026, 602 organizations

36%

of owners have a written AI policy; six in ten say staff use public chatbots

Nationwide survey, September 2026

85%

of incurred cyber losses in the first half of 2026 came from attacks on people, not systems

Resilience, July 2026

Those are the three questions a managed service already answers for your devices, identity and backups: what is in use, whether it is safe, whether it pays off. For AI, nobody is answering them. Strategy firms hand you a deck; the tool vendors sell seats. The work that decides an AI program happens on day two, in access reviews, vendor renewals, spend caps and audit evidence, and that is MSP work.

Identity is the door AI walks through. We already hold the keys for our clients: Microsoft 365, Entra, Conditional Access, the endpoints and the backups. Gating an AI tool behind them is a configuration, not a project. Doing it without them is where companies get hurt.

What a Managed Intelligence Provider does.

A Managed Intelligence Provider is an IT partner that does everything a managed service provider does and manages the client’s AI footprint with it. The term was coined by Pax8 in June 2025, written into its MIP Playbook that October, and taken up by MSPs, vendors and, in October 2026, a GTIA alliance writing standards for it. The definition we use is Portal26’s: “what’s being used, whether it’s safe, whether it’s paying off.”

The label has critics, and we agree with them. Dave Sobel of Business of Tech called it “a hard no… the endgame isn’t ‘intelligence’; it’s outcomes.” The label is the channel’s. The service is what the monthly report shows.

A managed service providerDevices, identity, network, cloud, backups. A monthly retainer per user. Owns day two for IT. Produces patch, backup and ticket reports.
An AI consultancyStrategy, pilots, sometimes a build. A project, then it leaves. Owns nothing after the handover. Produces a deck.
A Managed Intelligence ProviderThe AI estate on top of the IT estate. A retainer per user and per workflow. Owns day two for AI and IT together. Produces a monthly report on use, risk and return.
What it needs to workTooling, a help desk and identity already in hand, which is why it comes from an MSP and not from a strategy firm or a software vendor.

Five services, one retainer.

Four of them are the retainer: ongoing work, reported every month. The fifth is project work with a start, an end and a deliverable, quoted before it begins.

KnowA register of every tool, agent and subscription, found from sign-in logs, web filtering and SaaS discovery, with an acceptable-use policy your staff can read in five minutes.In the retainer · per user, monthly
SecureMFA, SSO and Conditional Access in front of every approved tool; enterprise data terms, BAAs where needed, audit logs and data-loss controls behind it.In the retainer · per user, monthly
UseRole-based training for executives, managers and frontline staff, the prompt library and office hours, so the tools get used, and used well.In the retainer · per user, monthly
RunMonitoring, spend caps, incident response and vendor changes for the agents and workflows you come to depend on, each with a runbook.In the retainer · per workflow in service, monthly
MakeWorkflows and agents built on n8n and Claude, one pilot at a time, with your team in the room and a measure agreed before work starts. See AI & Automation.Project work · fixed fee per build, from $3,500

What that means: the retainer is one monthly fee for the work that never ends and the report that proves it, on 90-day terms that renew on their own. A project is priced and agreed before it starts and ends with something you own. When a build goes live, the retainer takes it over under Run, which is why Run is billed per workflow.

How an engagement starts.

Audit, pilot, then managed. Stop after any step and keep what you paid for: the report, the policy, the pilot. Nothing is held back to force the next one.

01 / Weeks 1 to 2

AI footprint audit

What is in use, who has access, what data moves, what it costs. A written report, a prioritized list and a draft policy.

$2,500, credited to the next step

02 / Weeks 3 to 13

90-day pilot

Policy, identity gating and one running workflow with trained users. Four phases, one deliverable each; the AI onboarding plan.

Scoped from the audit

03 / Monthly

Managed Intelligence

The five services run as one retainer with a monthly report. Ninety-day terms that renew on their own; smallest client, one user.

Per user, plus per workflow in service

The monthly report.

One report that answers the three questions, written to be handed to an insurer, an auditor or a board without a separate project.

What is in useRegister changes, new sign-ups caught, seats and licences, the agents and workflows in service. Read by the owner and the operations lead.
Is it safeAccess reviews done, incidents and near-misses, policy exceptions, evidence ready for insurers and auditors. Read by the owner and whoever holds compliance.
Is it paying offSpend by tool and workflow against the caps, hours returned on the pilots we measure, next quarter’s pilot. Read by the owner and finance.
Why it matters nowTexas has made a documented AI governance program a legal defense since January 2026, auditors ask for AI in SOC 2 system descriptions, and carriers are writing AI into and out of cover. The report is the evidence pack for all three.

What the rules, the insurers and the incidents now say.

Ohio has no private-sector AI statute yet. Clients, applicants and customers elsewhere bring these in, and most small firms are inside at least one of them through hiring.

Rules, by effective date

  • October 1, 2025California’s automated-decision rules in employment: four years of records for any AI used in hiring or other employment decisions.
  • January 1, 2026Texas Responsible AI Governance Act, for anyone serving Texas residents; a documented NIST AI RMF program is an affirmative defense. Illinois HB 3773: notify people when AI is used in an employment decision.
  • August 2, 2026EU AI Act, Article 50: transparency duties reach US firms whose AI output is used in the EU.
  • January 1, 2027Colorado SB 26-189 and California’s ADMT rules: notice when AI materially influences a decision about a person, a trained human with override authority, records kept. California’s rules reach only CCPA-covered businesses.

What insurers are seeing

Business-email compromise and funds-transfer fraud were 58% of cyber claims in Coalition’s 2026 claims report. Resilience put 85.3% of incurred losses in the first half of 2026 on attacks against people rather than systems, up from 17.7% two years earlier. The FBI logged $893 million in losses from 22,364 AI-enabled fraud complaints in 2025. Cover is splitting: Coalition sells an affirmative AI endorsement and deepfake-response cover, while Berkley added an absolute AI exclusion to its D&O and E&O forms and ISO is filing generative-AI exclusions for general liability. No carrier publishes its AI questionnaire yet; what an underwriter can ask for is what the monthly report holds. Our cyber-insurance readiness checklist covers the ten controls the AI tools now sit outside of.

Incidents, August 2025 to October 2026

  • August 2025. Stolen Salesloft Drift OAuth tokens were used to export data from many companies’ Salesforce instances. Every integration token is an identity.
  • January 2026. ShareLeak: text typed into a public SharePoint form made a Copilot Studio agent email customer data to an attacker. A public form is an input to your agent.
  • April 2026. An AI coding agent wiped PocketOS’s production database and its backups in nine seconds; the car-rental firms on its software lost months of bookings. Agents need scoped credentials, backups they cannot reach, and a person who can stop them.
  • June 2026. Attackers talked Meta’s AI support chatbot into resetting Instagram logins without checking identity. A bot that can reset access is a target.
  • June 2026. SearchLeak: one click on a microsoft.com link made Microsoft 365 Copilot search a mailbox and leak it, rated critical. Links are the new attachments.
  • October 2026. AI agents run by one hacker broke into more than 100 company sites for about $5,000 in tokens. Attackers run agents too; MFA, patching and an inventory are the baseline, not the plan.

What we will not do.

The hard lines that keep the work honest, carried into every proposal.

We will not train foundation models for you.That is a different business and a different cost structure. If you need a custom model trained from scratch, we will tell you so and refer you.
We will not build a customer-facing chatbot as a first project.The data, the risk and the brand exposure are wrong for a start. Third or fourth project, once the muscle exists.
We will not deploy AI into environments without MFA on identity.If you have not enforced MFA, that is the first project, and it is not an AI one.
We will not advocate for a single vendor.Recommendations rotate with your environment, regulatory scope and use case. No resale relationship distorts them, and AI usage passes through at cost under a cap you set.
We will not promise that an AI replaces a person.We measure the hours a workflow gives back and report them. Agent operations are sold one measured pilot at a time.
We will not put a number on a slide we cannot source.Every figure on this page and in the deck carries its publisher and date, and the ones we could not trace stayed off.

For partners that build AI engines.

Outbound agents, revenue-intelligence tools and custom builds land in a client environment that then needs an owner. The engine needs aligned SPF, DKIM and DMARC and a spam-complaint rate under 0.3%, because Google, Yahoo and Outlook now reject bulk mail that lacks them; prior express consent before any AI-voice call; least-privilege CRM and mailbox access through named service accounts; audit logging, a spend cap and a kill switch for every agent; and someone to hand the evidence to an insurer or auditor. The failure mode is documented: in August 2025, stolen integration tokens were enough to export Salesforce data from many companies at once.

Our proposal to the firms that build these engines is a two-gate model: you certify the list and the message, we certify the environment, and the environment stays on the retainer after the campaign launches.

ReferYou refer, we deliver under our own agreement. Your client keeps one engine and one operator. Our referral program pays no fees in either direction.
Co-deliverA secure AI landing zone before your agents go live: identity, consent, scopes, agent identity, logging, caps. We run it afterwards.
PortfolioManaged Intelligence sold through your engine as a line in your portfolio, with the proposal kit, the deck and the monthly report ready to carry your name beside ours.

Common questions

An IT partner that does everything a managed service provider does and manages the client’s AI footprint with it: what is being used, whether it is safe, and whether it is paying off. The term was coined by Pax8 in June 2025 and has since been adopted by MSPs, vendors and the GTIA trade association. We use it because clients recognize it; the service is the outcomes on the monthly report, not the label.

Two weeks: every AI tool, agent and subscription in use, found from sign-in logs, web filtering and SaaS discovery; who has access to each; what data classes reach them; what they cost; and the identity and MFA posture they sit behind. You get a written report, a prioritized list and a draft acceptable-use policy. It costs $2,500 and is credited in full to the pilot or the first three months of the retainer.

It combines published lines: Managed IT from $125 per user per month, Cybersecurity from $45 per user per month, and AI and automation builds from $3,500 per workflow. The AI governance add-on is priced per user after the audit, and workflows in service are priced per workflow per month. Terms are a monthly retainer on 90-day terms that renew on their own; the smallest client is one user. For comparison, published US MSP prices in 2026 run $2,500 to $7,000 for an AI readiness assessment and $750 to $2,500 a month for an AI governance retainer. The full table is on pricing.

No. The audit and the pilot stand alone, and the retainer can sit beside your existing IT provider. It works best when we also hold identity, because MFA, SSO and Conditional Access are how every AI tool gets gated; if another provider holds them, we coordinate with them on access controls.

None in particular. Recommendations rotate with your environment, regulatory scope and use case, across Microsoft Copilot, Claude, ChatGPT Enterprise, Gemini and the tools your staff already use. We hold no resale relationships that distort that, and AI subscriptions and API usage pass through at cost under a cap you set.

Yes. The PDF on this page is the prospect deck we present, with every figure dated and sourced on the slide. The monthly report an engagement produces is written to be handed to an insurer, an auditor or a board without a separate project.

Twenty minutes of people talking to people.

We look at what is in use today and tell you whether an audit is worth your money. No pressure, no pitch. A straight answer either way.