Deep Research in Action

A real project, run on Delvantic. Three iterations, each one deeper than the last. Every output file is yours to read.

SaaS Market Viability Study 3-Part Series

Should We Build This? Evaluating an AI Voice Agent Startup from Brand to Pivot

The Scenario: Two technical co-founders built an AI voice agent SaaS for home service businesses — HVAC, plumbing, electrical, roofing. The product worked. Customer acquisition didn't. Like most early-stage entrepreneurs, they hit the question every founder eventually faces: Is the market the problem, or are we the problem? Rather than guessing, they decided to run the numbers — and keep digging until they had a real answer.

This is what iterative deep research looks like. The first run cast a wide net. The entrepreneur reviewed the output, decided what mattered and what didn't, then refined the direction for V2. Each version goes deeper — not by asking the same question again, but by steering the research based on what you learned. That's the core of Delvantic: you delve. Three runs. $11.01 total. Under an hour. Click any file below to read the actual output.

V1 Completed

The Wide Net — Competitive Landscape & Brand Positioning

The first run cast a broad net across the problem space: competitive brand audit of 11 players (Avoca AI, Goodcall, Synthflow, Smith.ai, and more), naming analysis, visual identity exploration, tagline candidates, landing page blueprint, and an AI hype strategy analysis. It produced 7 structured deliverables covering everything from market positioning to color palettes.

00-EXECUTIVE-SUMMARY.md naming-analysis.md visual-identity.md competitive-brand-audit.md taglines-and-copy.md landing-page-narrative.md ai-hype-strategy.md

Opus 4.6 14 min $1.76 7 files 37 turns

The Pivot Moment

V1 delivered useful competitive intelligence — 11 competitors mapped, market dynamics surfaced, early warning signs identified. But it also spent significant effort on branding, naming, and visual identity that the entrepreneur didn't need. This is exactly how iterative research works: V1 reveals the landscape, and you decide where to drill down. The founder cloned the project, stripped out everything branding-related, and steered V2 toward pure financial analysis.

V2 Completed Refined from V1

Going Deeper — Hard Numbers & Market Viability

After reviewing V1, the entrepreneur redirected: "No branding, no colors, no slogans. I want the data." V2 was a pure financial and competitive intelligence exercise. Comprehensive census of 35+ AI voice agent companies with funding, pricing, and headcount. TAM/SAM/SOM market sizing with real math. Unit economics modeling — CAC, LTV, payback period. Market saturation projections. And a brutally honest go-or-no-go verdict.

00-EXECUTIVE-SUMMARY.md competitor-census.md market-sizing.md unit-economics.md saturation-analysis.md market-dynamics.md go-or-no-go.md

Opus 4.6 17 min $2.46 7 files 48 turns

Verdict: NO-GO

The market exploded from ~10 competitors in 2022 to 35+ in 2025 with $200M+ in VC funding, while prices collapsed 61%. Worse: Jobber (200K customers), Housecall Pro, and CallRail all added AI voice as a free feature — meaning they acquire home service customers at zero incremental cost while a bootstrapped team spends $350-500 per customer. Probability-weighted 12-month outcome: ~$40K ARR total. Below minimum wage for two people.

V3 Completed Redirected from V2

Deepest Level — If Not This, Then What?

V2 delivered the hard truth: this market is a no-go. Instead of stopping there, the entrepreneur redirected again: "The voice AI space is oversaturated. Find me underserved markets with high-paying customers and low competition." V3 scanned 20+ industries, identified gaps where AI could create real value, built a longlist of 15-25 opportunities, then did deep dives on the top 5 with full unit economics, risk scoring, and a concrete 90-day action plan for the winner.

00-EXECUTIVE-SUMMARY.md opportunity-longlist.md top-5-deep-dives.md market-gap-analysis.md unit-economics-comparison.md risk-matrix.md action-plan.md

Opus 4.6 25 min $6.79 7 files 47 turns

A Clear Path Forward

Of 20+ opportunities analyzed across industries like legal, healthcare, construction, and professional services, V3 identified a specific underserved vertical with an 18:1 LTV:CAC ratio, a 2-month payback period, and zero vertical-specific AI competitors. The research surfaced a defensible moat that well-funded generalist platforms won't bother replicating — plus a concrete 90-day action plan with week-by-week milestones and kill criteria.

This Is What "Delving" Looks Like

Three runs. Each one sharper than the last — not because the AI got smarter, but because the human steered it. V1 was the foundation: broad, exploratory, some useful findings mixed with things the entrepreneur didn't need. That's normal. The first run is never the final answer — it's the starting point.

V2 went deeper because the founder knew what to cut and what to amplify. No branding. No slogans. Just data. The result was a financial verdict backed by real competitor counts, real pricing data, and real unit economics. V3 took that verdict and ran with it in an entirely new direction — scanning for overlooked markets and producing a ranked list of alternatives with a full action plan.

This is how Delvantic is designed to work. You don't get your answer on the first try — you get it on the third, or the fourth, or the fifth. Each iteration costs a few dollars and takes minutes. By the end, you have something that would take a research agency weeks and tens of thousands of dollars to produce.

Read through the output files above. Click any filename. See for yourself what each level of depth produced — and imagine what you could uncover with your own question.