How Long Does an AI Project Take: 2 Hours-2 Weeks

The short answer to “how long does an AI project take” is: generally 2 hours to 2 weeks. Connecting a WhatsApp bot to a system finishes in a few hours; setting up a 43-account Google Ads reporting automation stretches to two weeks. What determines the timeline isn’t how much “AI” is in the job, it’s how clearly the scope is defined. The 8-16-week estimates you often hear in the industry are the timeline of the large-agency model — for small, well-defined jobs, you don’t need to wait that long.

Why Some Jobs Finish in 2 Hours and Others Take 2 Weeks

Both ends are real: some jobs finish in an afternoon, some stretch across two weeks. The difference isn’t how advanced the AI is — it’s how narrowly and clearly the scope of the job is defined.

2-4 hours: connecting a single channel. Connecting a bot to a WhatsApp line, setting up an auto-reply on an email inbox, preparing a custom reporting flow for a single account — all single-system, single-integration, clearly defined tasks.

1-3 days: jobs requiring multiple systems to talk to each other. A setup matching a CRM with a calendar and email; a custom GPT/assistant fed by company data; an automation that produces a single report from multiple data sources.

1-2 weeks: where scale itself grows the job. Setting up a 43-account Google Ads reporting system isn’t technically different from a single account — but it needs to be tested 43 times, needs 43 separate access grants defined, and needs 43 potentially different exceptions handled. The job doesn’t get harder, it multiplies.

What Actually Stretches the Timeline: Not Scope, Waiting

The technical part is rarely the real bottleneck. The real reason a 2-day job jumps to 2 weeks is usually one of three things:

• Waiting for approval. A client who sends an email and gets a reply 3 days later stretches a technically 1-hour job into a 1-week calendar slot. That’s not the duration of the work, it’s the speed of communication.
• Integration surprises. A system’s API might not work as documented, an old piece of software’s access permissions might turn out harder to obtain than expected. These are unpredictable but not rare.
• Data quality. If a CRM’s name fields are inconsistent, if an Excel’s date format is a mess, the data needs to be cleaned before the automation itself can even start. This step usually takes longer than estimated.

I don’t hide these because hiding them helps no one — I state all three risks upfront when giving a time estimate.

Why “8-16 Weeks” Is the Industry Norm, and Why My Model Is Different

It’s not surprising to get an 8-16 week proposal from an AI/automation agency — but that’s a function of that model’s structure, not the nature of the work. In the large-agency model, a project manager collects the scope, a business analyst extracts requirements, a dev team writes the code, a QA team verifies it, and weekly status meetings and approval chains run between all of them. Every handoff adds a waiting period.

I’m one person: I hear the scope, I write the code, I test it, I decide in the same call. No handoffs, no status meetings. This isn’t a “better” model — it’s a faster one for small-to-medium, well-defined jobs. For very large, multi-stakeholder enterprise projects, the agency model has its reasons; that needs a different structure.

An Observation From the Field

Two examples, in the same week, showing the difference clearly. For one client, a WhatsApp-connected Q&A bot was built and tested in 3 hours — single channel, single task, clear scope. For another client, a 43-account Google Ads reporting system stretched across 10 days; the technical work itself took 2-3 days, the rest of the time went into waiting for each account’s access approval.

Same week, same consultant, two jobs ranging from 3 hours to 10 days. The difference wasn’t skill, it was scope.

How the Time Estimate Is Given

In the first call (generally 30-45 minutes, remote) we clarify the scope of the work together: which systems are involved, who will give approval, what data will be used. I give a concrete time estimate either before the call ends or the next day — “this job is 4-6 hours,” “this job is 5-8 business days.” Since billing is hourly, the estimate isn’t a hard ceiling, it’s a realistic range; you’re billed for however long it actually takes me to finish. For details on how pricing is determined, see AI Consulting Pricing.

Honestly, not every job fits into this range. Large integrations requiring many systems to connect simultaneously, projects requiring training a model from scratch, or very large data migrations don’t fit this model — those need a different way of working and a different timeline. This page covers small-to-medium, clearly defined jobs; if the scope turns out to be large during the call, I say so upfront — I don’t understate the timeline and stretch it later.

You Can Ask Me to Do This

The scope and timeline logic above is the exact description of the work I take on: generally 2 hours to 2 weeks, entirely remote, hourly rate. Describe your scope, and I’ll give you a clear timeline and cost estimate in the same call. See the AI Consulting page for service details.

Frequently Asked Questions

Does a 2-hour job really finish in just 2 hours?

If the scope is genuinely narrow and single-system, yes. For example, connecting a predefined bot to a WhatsApp line, or setting up an auto-reply on an email inbox, including testing, can finish in 2-4 hours. How clearly defined the scope is is the single variable that determines the timeline.

Which type of job takes the longest?

Multi-account or multi-branch structures — for example, a reporting/automation system covering dozens of accounts. The technical work itself isn’t large, but each account means separate access, separate testing, separate potential exceptions — which is why it can stretch to 1-2 weeks.

What determines the time estimate?

Three things: how clearly the scope is defined, how many separate systems are touched, and how fast approvals come through. Once these three are clear in the first call, the estimate becomes clear too.

What happens if my job looks like it will take longer than 2 weeks?

I say so clearly in the first call. Very large integrations or building a model from scratch don’t fit into this 2 hours-2 weeks range; in that case I propose a different working plan — I don’t understate the timeline and stretch it later.