Key takeaways
- There are five realistic routes. Each is the right answer for some businesses.
- Your IT provider is good at identity and permissions and usually less good at changing how people work.
- Hourly consulting suits open-ended problems. A first AI rollout is not an open-ended problem.
- Very small teams with no sensitive data should usually do it themselves.
Once a business decides to use AI properly, the next question is who does the work. The market offers five broad answers, and the people selling each one tend to describe the others unfairly. We sell one of the five, so read this with that in mind. We have tried to say plainly when each of the others is the better choice.
The five routes at a glance
| Route | Typical cost | Speed | Best at | Weakest at |
|---|---|---|---|---|
| Do it yourself | Internal time only | Slow unless someone owns it | Small teams, low sensitivity | Follow-through, policy, training |
| Your IT provider (MSP) | Hourly or added to your retainer | Varies with their workload | Identity, SSO, permissions, devices | Use cases, model choice, adoption |
| Microsoft or Google partner | Project fee, often tied to licence sales | Moderate | Deep work inside one ecosystem | Neutral advice across vendors |
| Hourly AI consultant | Roughly 150 to 500 USD an hour, more for large firms | Depends on scope discipline | Open questions, custom builds | Cost certainty |
| Fixed-price implementer | Published package price | Fast, usually weeks | A defined, standard rollout | Bespoke engineering |
1. Do it yourself
For a team of under ten with no regulated or confidential client data, this is often the right answer. Pick a provider, buy business seats, turn on the obvious admin settings, write a one-page policy and spend an hour showing people what good use looks like. Our AI readiness checklist and usage policy guide cover most of it.
The failure mode is not technical. It is that the rollout becomes nobody's job, drifts for two quarters, and staff carry on with personal accounts. If you go this way, name one person and give them a 30 day deadline.
2. Your IT provider
Your managed service provider already knows your tenant, your identity setup and your devices. For single sign-on, permissions clean-up, conditional access and licence assignment, they are often the best people to do the work. We cover this route in more depth in AI consultant vs your IT provider.
Where MSPs tend to be thinner is everything that is not infrastructure: choosing between providers on the work rather than the licence, matching models to tasks, writing a usage policy people follow, and training staff on their actual jobs. Ask yours what they have done on those parts before handing them the whole thing.
3. A Microsoft or Google partner
If you have already decided on Copilot or Gemini and the work is deep inside that ecosystem, such as Copilot Studio agents, Power Automate flows or SharePoint restructuring, a specialist partner is a strong choice. They will know the platform better than a generalist.
The trade-off is neutrality. A partner whose revenue is tied to one vendor will rarely tell you that a different tool suits a particular team better. If you have not yet chosen a provider, get that decision made independently first.
4. An hourly AI consultant
Hourly consulting is the right structure when the problem is genuinely open: you do not know what you want to build, the work involves custom software, or the scope will only emerge as you go. Good consultants are worth their rate for that kind of work.
For a first rollout of business AI assistants, though, the scope is knowable in advance. Provider, tier, admin settings, spend caps, policy, training. Paying hourly for knowable work puts the estimation risk on you. See fixed-price or hourly AI consulting for the longer argument.
5. A fixed-price implementer
This is what we do, so weigh this section accordingly. A fixed-price implementer sells a defined package at a published price: provider chosen and configured, spend capped, admin and SSO set, staff trained, done within a fixed period. You know the cost and the end date before you start.
It is the wrong choice if what you need is bespoke engineering, such as a custom application built on a model API, or a long programme of process redesign across many departments. It is a good choice for getting a governed, capped, well-adopted assistant rollout done quickly.
How to choose in five questions
- How many people, and how sensitive is your data? Under ten with low sensitivity: do it yourself. Above that, or with client confidentiality obligations, get help.
- Have you chosen a provider? If not, get neutral advice before talking to a vendor partner.
- Is the work standard or bespoke? Standard rollout: fixed price. Custom build: consultant or developer.
- Who will handle identity and permissions? Often your MSP, whoever leads the rest.
- Who owns it after go-live? If the answer is nobody, fix that first, whichever route you take. Read the first 90 days after go-live.
If you want the named-company version of this comparison, read the best AI implementation companies for small business, where we rank specific firms against published criteria.
Frequently asked questions
Should a small business hire an AI consultant?
Only if the value of getting it right quickly is larger than the fee. For a team under ten with no sensitive client data, doing it yourself with a good checklist is often enough. Above about fifteen people, or with confidential client data, outside help usually pays for itself through faster adoption and fewer mistakes.
Can our IT provider set up AI for us?
Your IT provider is well placed for identity, single sign-on and permissions, and less often for choosing models, writing usage policy and training staff to use AI in their actual work. Many businesses get the best result by having an AI specialist lead the rollout and the IT provider handle the identity pieces.
How much does it cost to have AI implemented?
Published market prices range from a few thousand dollars for a fixed-scope setup to tens of thousands for consulting-led programmes. Hourly consulting typically runs from 150 to several hundred US dollars an hour. A fixed fee is better for a first rollout because the scope is knowable.
What is the cheapest way to implement AI properly?
Buy business-tier seats from one provider, configure the admin settings and spend caps yourself using the provider's documentation, write a one-page usage policy, and run a single training session. It costs internal time rather than fees. The risk is that it drags and nobody owns it.
Get a straight answer for your business
The Clarity Package gives you a written business case and a detailed implementation plan for 500 USD. The Implementation Package does the whole build, live in 30 days, for a fixed 5,000 USD.