What happens when an IT Service Provider tries to create, send, and close a simple quote using AI — and why “paving cow paths” is the perfect metaphor.
The Simple Challenge
We set out to answer a deceptively simple question:
Can AI actually help an IT service provider close a deal, or just make a nice-looking document?
The task:
Create and send a professional-looking quote for cloud backup protection to a potential client — without major editing — in a way that could flow directly into fulfillment systems.
We tested five AI approaches:
- Claude (general AI with research)
- ChatGPT (general AI, no research tools)
- Microsoft Copilot (embedded business AI)
- Google Gemini (consumer flagship AI)
- Zomentum (industry-specific quoting tool)
The Quick Verdict
AI performed one step in the quote-to-close process: document creation — roughly 5% of the total work.
The other 95% — relationship building, strategic positioning, execution, delivery — still required humans, exactly as before.
The Results at a Glance
| Tool & Type | Output Quality | Context Understanding | Workflow Integration | Risks / Limitations | Human Work After “Yes” |
|---|---|---|---|---|---|
| Claude (general AI + research) | Well-formatted, accurate pricing; no branding | High — understood client already had service | None | Needs branding & manual system entry | 95%: CRM, contracts, onboarding, delivery |
| ChatGPT (general AI, no research) | Simple, accurate, clean template | Good industry context | None | No system connections | 95%: Same as Claude |
| Microsoft Copilot (embedded) | Professional doc from real internal data | Medium — mined existing emails well, but no added intelligence | None beyond Office apps | Results not repeatable without same data access | 95%: Same as Claude |
| Google Gemini (consumer AI) | Detailed specs, inflated pricing | Low — misunderstood business relationship | None | Pricing 2–3× market rate; wrong partner role | 100%: Quote unusable without rewrite |
| Zomentum (industry-specific) | Professional, SLA included, integrated with quoting workflows | High — positioned as service transition | Partial — connects to some CRM/fulfillment | Fabricated SLA terms & fees; legal risk | 90%: Oversight, corrections, relationship mgmt |
What Happens After ‘Yes’
Even the best AI output still required:
- Adding branding & contact info
- Entering into CRM/quoting systems
- Generating proper contracts & terms
- Billing & service delivery setup
- Onboarding, relationship management, support
In other words: AI helped with Step 3 of 8.
The Enterprise Barrier
We also attempted to test Salesforce AI. The experience revealed a common trap:
By the time you’ve configured the infrastructure needed to test the AI, you no longer need the AI for that task.
For small teams, this “setup before value” problem makes many enterprise AI tools impractical — the cost of evaluation can exceed the cost of solving the problem manually.
Key Insights
- Industry-specific ≠ better — Zomentum integrated with workflows but hallucinated legal terms.
- Data access ≠ business value — Copilot mined data but produced no new insights.
- General-purpose can outperform specialized — Claude & ChatGPT were more reliable.
- Context understanding is inconsistent — Gemini misread the entire relationship.
- The last mile is always human — Closing deals still depends on human execution.
Conclusion — Still Paving Cow Paths
AI has become very good at automating the least valuable part of the sales process — formatting a quote. But until AI can:
- Manage relationships
- Handle contract execution
- Orchestrate service delivery end-to-end
…it will remain an expensive way to pave cow paths, not build new highways.
The real question isn’t “Can AI create a quote?” — it’s “Can AI close deals and deliver value?”
Across all five tools, the answer is still: Not yet.