~/discovery $ cat methodology.md

We don't guess at AI opportunities. We measure them.

Most AI consulting firms start with workshops and interviews. SkyView Labs starts with data. Our discovery methodology deploys lightweight behavioral telemetry agents across your workforce for 14 days, then runs an AI analysis pipeline over the results to produce prioritized, evidence-backed automation and AI opportunities — with specific estimates attached to each one.

Engagement length

14 days

Output

Ranked findings · hours-wasted estimates

Data handling

Private by default · air-gap available

Analysis engine

Self-hosted LLM · no external API calls

~/discovery $ cat how-it-works.md

How it works

SkyView deploys a signed, lightweight Windows agent (MSI installer, local admin, 5 minutes to deploy) to each participating machine. The agent runs as a background service — imperceptible to users, consuming less than 0.5% CPU and under 25MB of RAM. It captures behavioral signals only: which applications are in use, how often people switch between them, what file types they're handling, and where they spend their time.

Every 15 minutes, the agent batches and uploads its buffer — compressed, encrypted, HMAC-signed — to the ingest endpoint. After 14 days (or earlier if the engagement scope is met), the agent automatically stops capture and self-uninstalls. No manual removal step. No residual software.

At the end of the capture period, our AI analysis pipeline processes the aggregated behavioral profiles and produces a structured findings report: ranked opportunities, the telemetry evidence behind each one, estimated hours wasted per week, and a recommended approach — whether that's an RPA automation, a custom AI application, a system integration, or a process redesign. Each finding maps directly to a SkyView service.

// opa.skyviewlabs.com — consultant dashboard

OPA consultant dashboard showing engagement overview for ACME Financial Services — 94 of 100 machines reporting, 2.41M events captured, 41.5 hours wasted per week identified, 8 automation opportunities ranked by impact — The OPA consultant dashboard mid-engagement. Live event throughput, per-machine coverage, AI analysis runs, and ranked findings — all on SkyView's private infrastructure.

~/discovery $ cat captured.md

What we capture — and what we never touch

captured.txt

Foreground application name and window title (sanitized — see below)
Application transition events — what users switch from and to, and how long before switching back
Browser domain only (e.g., salesforce.com) — no full URLs, no page content
File operations by type: open, save, print — extension and operation only, never filename or content
Keyboard and mouse activity rate (events per minute — not keystrokes, not content)
Meeting presence — camera/mic active flag only, no audio or video
Idle and lock-screen events

never-captured.txt

Keystroke content — what you type is never recorded
Clipboard content — copy/paste data is never captured
Screenshots or screen recordings
File content — documents, spreadsheets, emails remain untouched
Email or chat message body
Passwords or form field content
Network traffic or packet data
Audio or video

// title sanitization Window titles are scrubbed before leaving the machine: email addresses, phone numbers, dollar amounts, and SSN patterns are stripped automatically. Application-specific rules apply additional redaction — Outlook strips email subjects, Chrome keeps only the domain.

~/discovery $ cat data-residency.md

Your data stays where you need it

Three deployment modes. Every engagement uses whichever fits your compliance posture. The capture agents are identical across all three — only the ingest endpoint changes.

deploy/standard.md

STANDARD

SkyView private server

Telemetry is uploaded encrypted and HMAC-signed to our self-hosted server — not a cloud provider, not a SaaS vendor. No third party ever receives your data. The analysis runs on a locally-hosted AI model. No telemetry ever leaves to an external AI API.

Data at: SkyView private cloud
AI model: Self-hosted — no API calls
Right for: Most mid-market engagements

deploy/compliance-mode.md

COMPLIANCE MODE

Stays in your network

A compliance node runs entirely inside your perimeter on a server you control. Raw telemetry never leaves your network. The AI analysis runs on-premises. SkyView receives only the final findings JSON (~50KB) — or nothing at all in fully air-gapped mode.

Raw data at: Your network, your hardware
AI model: On-prem, CPU-only, no GPU required
Right for: Healthcare · finance · legal · federal

deploy/air-gapped.md

AIR-GAPPED

Nothing leaves at all

In fully air-gapped mode, the compliance node is completely isolated. SkyView VPNs in (or visits on-site) to view findings and export the report. Zero data is transmitted upstream at any point. At engagement end, the Docker stack is torn down and all data destroyed on-premises.

Data transmitted: Zero — nothing upstream
Access: SkyView VPN or on-site visit
Right for: ITAR · air-gap mandates · sovereign

~/discovery $ cat findings-output.md

What you get at the end

The output is a ranked findings report — not a slide deck of generic AI suggestions, but specific opportunities backed by telemetry evidence with estimates attached.

finding-01.json

title Manual invoice reconciliation — high-frequency Excel/email loop

evidence 14 users averaging 2.3h/day in Excel → Outlook → Excel transitions, 09:00–11:30 daily

hours_wasted_pw 162 hrs/week across affected users

solution API Integration + AI document extraction

complexity Medium

priority_score 9.4 / 10

Each finding includes:

The specific behavioral pattern observed in the data
Which users (anonymized) are affected and how often
Estimated hours wasted per week — conservative, data-backed
Solution category: RPA, API integration, AI copilot, custom app, or process redesign
Recommended build approach
How a SkyView engagement would scope and deliver it

The report rolls up to an executive summary with a total estimated hours-saved figure. Findings are ranked by a composite of impact, affected-user count, and implementation complexity. You leave with a prioritized action list — not a to-do for a future strategy conversation.

~/discovery $ cat next-steps.md

How discovery leads to a build

The Discovery engagement is flat-fee — typically $10,000–$25,000 depending on workforce size and organizational complexity. The fee is fully credited against any SkyView build engagement that starts within 90 days.

At the end of 14 days, SkyView presents the findings to your leadership team. We walk through the ranked opportunities, answer questions about methodology and evidence, and give you an honest read on which opportunities are worth building first and what each would cost. If we're the right firm for the work, we scope it. If we're not, the report is yours to act on independently — the deliverable is written that way.

Most clients end up with a shortlist of two or three builds. We scope the highest-priority one in detail and begin development. The rest go into a roadmap. The data from the discovery period continues to be useful for months.

~/faq $ cat discovery-faq.md

Data-Driven Discovery — frequently asked questions

The questions we get most often, answered. If yours isn't here, ask it on a 30-minute call — we answer the awkward ones too.

Will employees know they're being monitored?

Yes — employee awareness and consent are required. Before any agents are deployed, SkyView provides a plain-language disclosure document that the client distributes to participating employees. It explains what is captured, what is not, how long the engagement runs, and when the agent self-removes. Deployments without employee notice are not supported.

What happens to the data after the engagement ends?

The capture agent stops automatically at the engagement end date and self-uninstalls — no manual removal required. Raw telemetry on the SkyView server is purged within 30 days of engagement close. In compliance-node mode, all data is destroyed on-premises when the Docker stack is torn down. Clients receive the findings report; the underlying telemetry is not retained.

Can you deploy this in a Citrix or virtual desktop environment?

Yes. The agent supports Windows Server / RDS session-level deployment and Citrix Virtual Apps. For environments where all sessions run on shared hosts, a local relay is deployed on the session host — all sessions route through it, reducing the network footprint to a single upstream connection regardless of session count.

Can the agent be blocked by our antivirus or EDR?

The agent is signed with an EV code signing certificate. SkyView provides a whitelist guide for CrowdStrike, SentinelOne, and Microsoft Defender for Endpoint, covering process names, file paths, and the destination hostname. In most environments, a brief IT admin step before deployment handles this cleanly.

What if some employees work remotely and some are in office?

Hybrid workforces are the common case. In relay mode, office workers upload through the on-premises relay (no per-machine internet egress); remote workers time out on the relay in 3 seconds and automatically fall back to the SkyView endpoint. When a remote worker returns to the office, the agent detects the relay on the next hourly probe and switches back. No manual configuration changes.

Do you use AI to analyze the telemetry — and where does that AI run?

Yes — a self-hosted large language model runs the analysis. By default this is a 14B-parameter quantized model running on SkyView's own GPU server. No telemetry is ever sent to OpenAI, Anthropic, or any other external AI API in the standard configuration. For clients who want higher-quality analysis on complex engagements, Claude API analysis is available with explicit consent — but the default is always local, always private.

How is this different from just doing stakeholder interviews?

Interviews surface what people think they do. Behavioral telemetry surfaces what they actually do. The two rarely match. Discovery via telemetry consistently finds high-value automation opportunities that go unmentioned in interviews — because they're so routine that nobody thinks to flag them — and avoids over-investing in AI for processes that turn out to be fast, low-friction, and not worth automating.

~/contact $ open

Start with data, not assumptions.

A 14-day discovery engagement tells you exactly where AI will move your numbers — and where it won't. Flat fee, credited against any build.

Book a discovery call See our services