AI consultant for Chicago businesses

FitSmall and mid-size businesses · Chicago and remote

• Owners who know AI could helpbut not where to start
• Teams pasting work into ChatGPTwithout approved tools or rules
• Operations stuck on copy-and-pastebetween email, forms and the CRM
• Front desks that miss callsafter hours and during the rush

AI consulting is the work of matching a business problem to the right AI tool, then putting it into daily use with clear rules. Most small businesses are early: the U.S. Census Bureau found 19.8% of U.S. businesses using AI as of May 2026, and use among firms under 20 employees did not change significantly from December 2025 to May 2026 while it rose among larger firms. Small teams lack the hours to pick a use case, set it up and check it. At 312, the person who recommends an automation builds it and maintains it.

What can workflow automation with AI handle?

Workflow automation connects your apps so a form, an email or a new record triggers the next steps without copying and pasting. An AI step lets the workflow read and write text: classify an inquiry, pull fields from an attachment, draft a reply for a person to approve.

Typical first builds: website leads summarized and routed with a draft reply, or an intake form that creates the client folder, CRM record and welcome email. Zapier is the easiest for an owner to read (it says it connects to more than 9,000 apps), Make handles branching well, and n8n can be self-hosted when data must stay in your environment.

Can you build a custom ChatGPT or Claude assistant?

Yes. A custom assistant is a chat assistant with written instructions, approved reference documents and rules for what it may answer, so staff get consistent answers from your own material.

Good candidates: an HR assistant that answers from the handbook, or a proposal assistant that drafts from past proposals. We build them in your company’s ChatGPT or Claude business accounts so you own them, and test each on real questions before rollout. On business plans, OpenAI and Anthropic both state they do not train on your inputs and outputs by default. Still choosing a platform? Read our ChatGPT vs Copilot vs Gemini vs Claude comparison.

Can AI work inside Salesforce or our CRM?

Yes, and it often pays off there first, because the CRM is where notes, follow-ups and data entry pile up. AI can summarize activity onto a record, draft follow-ups from deal notes and route inbound leads.

We decide which fields an AI step may write and log every change so a person can review it. For the platform work itself (health checks, Flow automation, clean-ups), see our Salesforce consulting in Chicago.

Can an AI voice or chat assistant answer our customers?

It can answer, book and hand off to a person, if it works from live data and has clear escalation rules. We run one in production: Sofia, the AI receptionist inside 312 Scheduling, answers barbershop and salon calls 24/7 in English and Spanish, switching languages mid-call.

Sofia checks real-time availability before she books, and every call she handles appears in the shop’s activity log. Two legal points shape every client voice project. The FCC ruled in February 2024 that AI-generated voices are “artificial” under the Telephone Consumer Protection Act, so outbound AI calls need the same consent as prerecorded ones. And in Illinois, an assistant that recognizes callers by voice collects biometric data under BIPA, which requires notice and a written release first.

What AI governance does a small business need?

Approved tools, rules for what data goes into them, named points where a person reviews AI output, and an owner for each automation.

Our checklist is the voluntary NIST AI Risk Management Framework, whose four functions (govern, map, measure, manage) scale down well to a 20-person firm. It ends up in your AI acceptable-use policy. If AI touches hiring, promotion or discipline, Illinois HB 3773 (in force since January 1, 2026) also applies.

Why is team training part of an AI consulting engagement?

Training is the adoption step. People who will live with a new workflow or assistant need to know how it works, how to check it and what to do when it is wrong, or they stop using it.

For a single build, that is a working session with its users. For broader skills, we run hands-on AI training for Chicago teams on your own ChatGPT, Copilot, Claude or Gemini accounts. 312 has delivered AI training for companies from small and mid-size businesses up to Fortune 500 firms, and its founder has led AI training for 1,200+ people across four regions; see the published training results and, for company-wide rollouts, corporate AI training.

1. Free first call. What is slow, error-prone or expensive today. We reply within one business day.
2. Discovery. Short interviews with the people doing the work and a look at the tools and data involved.
3. Ranked use cases and a written scope. Each use case scored on value, risk and effort. You pick where to start and get a written scope and quote.
4. Data and policy check. Which accounts, which data, which human review points; an interim policy if you have none.
5. Build and test. We build the first use case and test it on real inputs, or practice data where those are sensitive.
6. Rollout and training. Users learn it hands-on, with written instructions and a named owner on your side.
7. Measure and decide. A few weeks later we review the agreed measure, fix what needs fixing and choose the next use case.

When the first build is an internal tool or portal rather than a configured workflow, it typically takes 4–8 weeks. Timelines are estimates and subject to change based on scope.

Questions to ask any AI consultant

• What have you built that runs today?ask to see it
• Which of our ideas would you not do?a good consultant cuts the list
• Where will our data go?business accounts, named vendors, in writing
• Who owns what you build?it should live in your accounts
• What happens when it is wrong?review points, logs, a fallback to a person
• Who maintains it after launch?a named person

Red flags

• Guaranteed savings or revenuenobody knows before discovery
• “Replace your staff” pitchesassistants take the routine work
• Builds in the consultant’s accountsgone when the contract ends
• One tool for every problema reseller, not an advisor
• No mention of data or Illinois lawthe risk lands on you

The first red flag has a public example: in August 2025 the FTC sued Air AI, which sold “conversational AI” to small businesses with claims it could replace customer service staff and earn back its cost within days or months; the FTC alleged that buyers often did not earn what was promised and that refund guarantees were rarely honored.

We recommend the cheapest option that does the job. When custom is right, our web app development team builds it; when the question is bigger than AI, start with an IT roadmap.

Two lessons carry into client work. From Bosanski: put a person between AI output and the customer wherever a mistake would cost trust. From Sofia: an assistant must read the live system, not a copy that drifts.

What moves the quote: the number of use cases, configuration versus custom code, how many systems connect, and how much training the rollout needs. Support after launch can run on a flat monthly retainer.