neesh Inc.
Competitive IntelligenceCost ArchitectureAI Economics

The Four-Cent Briefing

We monitored 30 web pages for a week. Total AI cost: four cents. Here's what we found — and what it means for how businesses should think about intelligence.

We monitored 30 web pages for one week. Checked each one daily. Detected changes, scored significance, narrated what mattered. Total AI cost: four cents. Here’s what we found — and why it changes everything about competitive intelligence pricing.


The Setup

Thirty pages across five hypothetical competitors in a professional services industry. Not a cherry-picked sample — a realistic spread of the pages that matter when you’re watching a competitor’s posture.

The mix: pricing pages, team pages, service descriptions, blog and news sections, and career pages. One check per page per day for seven days. Two hundred and ten total checks. Every page, every day, no exceptions.

The goal wasn’t to detect every pixel change. The goal was to surface what mattered — and ignore everything that didn’t.


The Week

Most days, most pages don’t change. When they do change, most changes are cosmetic. That’s not a failure of the system — it’s the system working. Noise suppression is the job.

No change detected
Change detected, insignificant
Significant change narrated
Critical change requiring action
No change detected 72%
Change detected, insignificant 18%
Significant change narrated 7%
Critical change requiring action 3%

One hundred and fifty-one of two hundred and ten checks found nothing worth reporting. Thirty-eight found changes — new whitespace, formatting tweaks, a reordered FAQ — that scored below the significance threshold and were silently discarded. Fifteen changes were substantive enough to narrate. Six required immediate attention.

The ratio matters: for every hundred page checks, fewer than five produce intelligence worth acting on. Everything else is noise. The architecture’s value isn’t in detecting change — it’s in knowing which change to surface.


The Intelligence

The fifteen narrated changes varied in importance. Three stood out.


The Bill

Every step in the pipeline has a cost. Most of them are zero.

1

210 page checks

$0 — Playwright fetches pages. Open source, runs locally, no API cost.

2

210 diff comparisons

$0 — Text diffing is code. No model, no API, no charge.

3

210 ML significance scores

$0 — HuggingFace classifier runs locally. No inference cost.

4

21 LLM narrations

$0.005 each = $0.042 — Only substantive changes reach this layer.

5

1 weekly briefing

$0.01 — One compilation call to produce the formatted summary.

Two hundred and ten checks. Two hundred and ten diffs. Two hundred and ten significance scores. Zero dollars. The expensive layer — the LLM — fired twenty-one times. At half a cent per narration, the total inference cost was four point two cents. The weekly briefing compilation added one cent.

Total: $0.052. Round to the nearest memorable number: four cents.


What This Means

Competitive intelligence used to require budget. Enterprise CI tools — Crayon, Klue, Kompyte — price for the value they historically delivered: analyst hours, manual monitoring, bespoke research. That value was real. The pricing reflects it.

Enterprise CI tools
$1,250/month minimum, annual contract required, dedicated analyst time to interpret outputs, enterprise onboarding, vendor lock-in. Accessible only to businesses with five-figure annual intelligence budgets.
AI monitoring architecture
$0.17/month in AI inference costs, $35/month in infrastructure, no contract, briefings delivered automatically to Slack or email. Accessible to any business with thirty pages worth watching.

The filter-before-LLM architecture makes this possible. Ninety percent of the pipeline is free computation — Playwright, diff algorithms, a locally-run classifier. The LLM only fires when the cheaper layers have confirmed that something worth narrating has actually happened. The result is intelligence costs that don’t scale with volume — they scale with signal.

At four cents per week, competitive intelligence isn’t an enterprise budget line. It’s a utility.


Four cents. That’s not a loss leader or a promotional price. It’s the actual cost of monitoring 30 pages for a week when your architecture filters before it thinks. Intelligence shouldn’t be a luxury — it should be a utility.

Want this pointed at your competitors? Start with a free 30-minute assessment.

Book Free Assessment