Booking Q4 deliverystart with a free workflow plan Denver · Phoenix · Remote
The Field Guide / 10

ChatGPT subscriptions vs. custom AI: when does each win?

Start with the tool you already have. Build only when the workflow needs something it cannot do.

Field GuideNo. 10
Reading time3 min
SubscriptionsCheck the current plan
Custom build$18,000–$60,000

Is a ChatGPT subscription enough for your business?

It may be. Start by testing an approved business tool on the actual task, with the required permissions and a person checking the result. Buy a custom system only when a useful workflow needs connections, controls, or behavior the existing tools cannot provide.

The boundary is changing as subscriptions add apps, connectors, scheduled work, and agents. “A chat window cannot do that” is not a durable buying rule. Check the current product and plan before paying someone to rebuild a capability you already have.

Where should you try the existing tool first?

Try an attended task with a clear input and a reviewable output: a proposal draft, a summary of approved material, an internal checklist, or a response someone will edit before sending. Measure total time, including the check and any corrections.

Business tools may connect to company sources. For example, OpenAI describes business-workspace apps, permissions, and data controls. Availability and behavior depend on the plan and configuration. Review those settings before connecting sensitive material.

When does a workflow need more than a subscription?

When the tested product cannot meet a requirement that matters to the operation. Write the requirement in plain language before deciding what to build:

  • Trigger. What starts the work, and can the existing tool start it reliably?
  • Connections. Can it read and update the required systems with the right permissions?
  • Control. Can you separate routine actions from consequential decisions that need approval?
  • Evidence. Can your team inspect the sources, actions, failures, and review history it needs?
  • Handoff. Can someone maintain it, handle exceptions, and recover when a dependency fails?

A gap may call for a configuration change, an integration, or a defined build. It does not automatically call for a whole AI platform.

What does custom AI cost, and what do you get?

Our AI Implementation Sprint is $18,000–$60,000 over 4 to 12 weeks. The quote fixes the agreed scope before work begins. The deliverable is a working agent, automation, or integration tested in your operation, with documentation, ownership, and team training.

The scope defines what can run automatically and what a person must approve. Custom does not mean infallible or unattended. Budget tools and model usage separately; our typical planning band is $50–$500 a month, excluding hardware, internal time, the build, and optional Managed Services from $1,500 a month.

How do you compare the economics?

Use the actual plan price and the people who need it. For illustration, twenty seats at an assumed $25 a month cost $6,000 a year. That arithmetic is not a current quote for a named product. Check required licenses, billing commitment, usage limits, and taxes.

Then compare the same task under both options. Include setup, review, exceptions, support, and the amount of work actually removed. A cheaper license that leaves the whole handoff manual may be the wrong answer; a custom build for a task the subscription already handles may be wrong too.

How do you avoid buying shelfware?

Give one person one useful job and a short trial. Agree what a satisfactory output looks like, record the time before and after, and ask where corrections are still needed. Expand only when the evidence justifies it. Poor adoption can reveal a weak workflow, unsuitable software, missing permissions, or insufficient support; it does not have one universal cause.

Use the readiness checklist and ROI math guide to make that review concrete. Our calculator is a starting model, not measured savings in your operation.

What is the next step?

Bring the task and what you have already tried to the free 30-minute assessment. We will discuss whether to use the existing tool, investigate the workflow, or scope a build. If discovery is needed, an AI Readiness Audit is $3,500–$8,500 over 2 to 4 weeks. A defined workflow can go directly to sprint scoping.

Fair questions

Seats versus systems.

Want the numbers?

The full price list is published — audits, sprints, managed services, all on one page.

Read the price list →
01Is a ChatGPT subscription enough for a small business?+

It may be enough for individual drafting, research, and brainstorming. Test a defined workflow with approved data before buying a custom build. Repeatable work spanning several systems may need integrations, review controls, and monitoring.

02What does custom AI cost compared to ChatGPT seats?+

Compare the provider’s current seat price times users and billing period with the full custom workflow cost. Our Implementation Sprints cost $18,000–$60,000, plus running costs and any optional support; the two options deliver different scope.

03Why do ChatGPT rollouts fail in small businesses?+

Because seats get bought without any workflow being redesigned around them, so adoption fades within weeks and the licenses become shelfware. That outcome is evidence a change was never planned, not evidence AI failed.

04Should you try ChatGPT before hiring an AI consultant?+

A limited trial can reveal whether a subscription solves the workflow. Assign an owner, define approved data, and measure usage and output quality. Low adoption is a reason to investigate the workflow and training before investing further.

Start here

Seats problem, or system problem?

Bring us the workflow that keeps returning to your desk, and what you have already tried. In 30 minutes, we will discuss the existing-tool and scoped-build options. We reply within 24 hours.

We reply within 24 hours. A fixed quote follows an agreed scope, before paid work begins.

— Christopher Myers, Founder