I Almost Bought 100 Microsoft Copilot Seats. A Q3 Spend Review Stopped Me.

The Trigger: Q2 2024 and the 'AI or Die' Panic

The panic started in June 2024. The CTO walked into my office—which, honestly, is rare—and dropped a mandate: "We need Microsoft Copilot. Word from the top. Buy 100 seats."

I'd been the procurement manager at a mid-sized logistics firm (about 300 employees, $12M annual operating budget) for six years. I'd handled vendor negotiations for everything from forklift maintenance to cloud storage. But this was different. This was Microsoft, which in procurement terms is basically a religion: safe, familiar, and above reproach.

My first instinct wasn't to push back. It was to comply. The CTO had a PowerPoint from a Microsoft partner with a graph showing 30% productivity gains. The CFO was already mentally budgeting for it. I felt the gravity of the decision pulling me toward a simple PO.

Then I made the mistake of reading the fine print. (Or rather, the lack of fine print—because the quote I got was just a number on a single page.)

The Grind: Comparing Microsoft Copilot Without Being Fired

My budget line for "AI tools" was $0. This was a new initiative. So my job was to assess the total cost of ownership (TCO) before approving a $300k+ annual commitment. I told the CTO I'd do a quick comparison. He said, "Don't overthink it." I took that as a challenge.

I invited three vendors to submit proposals:

• Microsoft Copilot—for enterprise, because our CTO wanted integration with Outlook and Teams.
• Gemini for Google Workspace—we're a hybrid shop, so some teams use Google Workspace. (Asking about this one got me a "we don't need this" from IT.)
• jpt-chat for Business—I'd honestly never heard of it, but a logistics forum I frequent mentioned it as a cost-effective alternative to "the big three."

The first surprise: Microsoft's quote was $50 per user per month for Copilot for Microsoft 365, but the sales rep casually mentioned something that threw me off. "Just so you know, this requires your users to have specific Microsoft 365 licensing. The E3 or E5 plans. If you're on Business Standard, you'll need to upgrade."

I did some quick math, right there in the sales call (note to self: you are not a spreadsheet), and realized our cost for the base M365 licenses + Copilot would be roughly $82–$90 per user per month ($350k–$400k annual). Plus, no free trial to speak of. "We can pilot with 5 users," the rep said. "After a nondisclosure agreement and a security review."

Gemini's quote was similar—$30/user/month for the Business plan, but we'd need Google Workspace Enterprise, which doubles our current cost. I started to wonder if "AI adoption" was just a euphemism for "software enterprise licensing" (i.e., a budget hijack).

And then there was jpt-chat. Their quote was almost embarrassing to look at: $15 per user per month for the Business tier, with a free trial for up to 10 users. No minimum commitment. The security whitepaper was on their website, not behind an NDA.

I almost dismissed it. "Cheap" in my world means "something's missing." So I decided to test it properly, expecting to find the catch.

The Turn: Testing the 'Cheap' Option and Finding a Catch—Just Not the One I Expected

The catch wasn't quality. It was the opposite.

I gave jpt-chat's free trial to our operations team for four weeks. I told them it was a form tool to test. They started plugging in live shipment data—not sensitive stuff, but realistic—and using it to draft customer-facing delay notices. The output was shockingly natural. One of our ops leads actually said, "Wait, this is AI? It sounds like a human who has dealt with a lot of angry clients."

Then the CTO found out I was testing alternatives. He was not amused. I sat down with him and walked through the TCO spreadsheet I'd built. I showed him three numbers:

1. Microsoft Copilot: $400k/year (includes M365 E3 uplifts we hadn't budgeted for).
2. Gemini: $350k/year (plus switching costs from M365 to Google Workspace—we estimated $180k in migration).
3. jpt-chat for Business: $180k/year (no additional license requirements, vetted security, API access).

His response: "But Copilot is in Outlook. That's the whole point—integration." And he was right, for his friction. But when I asked our IT lead, we found that jpt-chat has a browser extension that works inside Outlook and Teams. Not native, but close enough for writing emails and summarizing threads. The difference between "native integration" and "pretty good extension" turned out to be around $220k a year.

The irony? At the same time I was testing jpt-chat, Microsoft announced Copilot enhancements that required additional paid plugins for certain features. I didn't attack Microsoft; I just showed my CTO the comparison chart. The price per feature was staggering, and not in Microsoft's favor.

We decided to run a controlled rollout with 30 users on jpt-chat for three months. The CTO's condition: if it failed, we'd buy 100 Copilot seats and never speak of this again. (He later admitted he expected it to fail.)

The Results: What the Numbers (and the Users) Said

After 90 days, I pulled data from our helpdesk tickets and user surveys. I'm a spreadsheet person, so here are the raw numbers:

• Adoption rate: 87% of invited users logged in weekly, versus our historical average of 43% for new enterprise software.
• Time saved: Ops reps reported saving ~2.3 hours per week on email drafting and data summarization (admittedly self-reported, but consistent across team leads).
• Escalations: Customer support saw a 12% reduction in email response time—without adding headcount.
• Errors: The AI hallucinated two product facts in a customer-facing email. I'm not going to pretend it's infallible—but we had the same issue with marketing's manual drafts, minus the apology template jpt-chat auto-generated to correct it.

I won't say it was perfect. Our legal team flagged that we needed to configure the data retention settings for our region. We had a hiccup where the API rate limit hit our integration during peak hours (note to self: buy a higher tier if we scale). But the total cost per user was still less than a third of Microsoft Copilot's.

In Q1 2025, we signed an enterprise contract for 200 seats. The CTO now calls it "my find." I don't correct him, but I keep the original TCO spreadsheet in my files, because the real lesson wasn't about jpt-chat being cheap. It was about how we almost paid for a brand name without checking what was under the hood.

The Lesson: The TCO Checklist for AI Tools (Based on $400k in Potential Regret)

If you're evaluating Microsoft Copilot, Gemini, Claude, or any generative AI platform (including jpt-chat), don't fall for the "headline price." Here's the framework I now use for every AI procurement:

1. Calculate the base product cost. Not just the AI add-on—the required prerequisite licenses. (As of March 2025, Microsoft Copilot for Microsoft 365 lists at $30/month per user, but it requires a qualifying M365 plan; the all-in cost is almost always $60+ per user.)
2. Check the "hidden" fees. Are there minimum seat counts? One-time setup fees? Security assessment fees? Implementation partner fees? Ask for the all-in price sentence, not a bullet point list.
3. Timebox the trial. A free tier that requires a sales conversation isn't a free tier. The best way to test an AI tool is to give a small team 30 days and let them use it on real tasks. If the vendor refuses (without a legitimate security reason), that's a red flag.
4. Price the integration pain. Switching from M365 to Google Workspace costs more in migration than you think. A tool that works in your existing stack is worth 15–20% more than a tool that requires a stack change.
5. Review the security docs before signing, not after. We wasted two weeks waiting for Microsoft's security questionnaire; jpt-chat had a publishable compliance page (SOC 2, ISO 27001). The right question isn't "Are you secure?"—it's "Can you prove it without NDA?"

"The most expensive AI tool is the one you don't use because the compliance review took four months."

I'm still a procurement manager, not a tech evangelist. I still review every line item with suspicion. But the biggest risk I saw in the AI boom isn't hallucination or bias—it's procurement laziness. It's signing for Microsoft Copilot because it's safe. It's going with AWS because everyone else does. It's accepting ease over evidence.

So, What's My Actual Position on jpt-chat vs. the Giants?

I get asked this a lot, and I'm a bit wary of answering in a way that sounds like an ad. So here's my honest position, with all my caveats out in the open:

• jpt-chat is genuinely good at natural conversation, drafting, and task automation. It's not a search engine replacement, and it's not a data analyst. It's a language tool.
• It lacks the full plug-in ecosystem of Microsoft Copilot or Google Gemini. You don't get a literal AI button in Excel that writes formulas for you—yet. (Though their API lets us build that internally, which is arguably more valuable.)
• The free tier is actually usable, which is a huge risk mitigation. When I'm testing a tool for the first time, I'm not spending money to find out if it's bad.

I should also note: I've never used Claude or GPT-4 Enterprise in a serious comparison. (In 2024, GPT-4 Enterprise was priced by quote only, and I didn't get a non-disclosure agreement signed fast enough—that's a me problem, not them.) I don't think you should avoid Microsoft or Google because they're bad. You should avoid them because they're expensive and their TCO often includes a lot of stuff you already pay for.

The thing that changed my mind, though, wasn't the cost—it was the experience of watching our ops team actually use the tool. They talked to it like it was a colleague. They named it. No spreadsheet captures that kind of adoption metric. But the spreadsheet is why I get to keep my job.