Implementation Blueprint
Overview
Most of the work in a Relyance rollout is not technical. Connecting a system is usually a read-only credential and a few clicks. The time goes into finding the right owner for each system and getting their approval.
Effort below means the work on your side to get a connection live:
- Low — one owner, one credential, done in a sitting.
- Medium — a few teams involved, or the same setup repeated across many accounts.
- High — real configuration and decisions, not just a connection.
| Integration Type | Typical Systems Involved | Typical Stakeholders | Level of Effort | Avg. Time Needed | Infrastructure Considerations (if any) |
|---|---|---|---|---|---|
| Source code | GitHub, GitLab, Bitbucket | Engineering leadership, DevOps/platform team, repo admins | Low — install the Relyance app at the org level and grant read access. Nothing to install on your side. | 1–2 days | Read-only access to repos. If you self-host GitLab or Bitbucket, Relyance needs a network path to it (IP allowlist or DirectConnect). |
| Cloud infrastructure (agentless) | AWS, GCP, Azure | Cloud/platform engineering, infrastructure security, IAM owners | Low to Medium — one read-only role per account. Easy for a handful of accounts; a rollout of its own if you have hundreds. | 1–2 days for a few accounts; 2–3 weeks for hundreds | No agents to install. Uses a cross-account role. Push the role out through your existing account-provisioning tooling rather than by hand. |
| Data storage and data warehouse | Snowflake, Databricks, ClickHouse, S3, BigQuery, RDS, PostgreSQL | Data engineering, DBAs, data platform owners, security | Medium — a read-only service account per system, plus agreement on when scans run. | 1–2 weeks per platform, mostly waiting on access approvals and a scan window | Classification reads real data, so it uses your warehouse compute — schedule scans off-peak and expect some query cost. Databases that are not publicly reachable need private networking (PrivateLink, VPC peering, or DirectConnect). |
| Enterprise AI | OpenAI, Anthropic, Bedrock, Vertex | AI/ML platform team, application engineering, AI governance and security | Low — an admin-level API key per provider. Bedrock and Vertex usually need nothing new, since they are already covered by the cloud role. | 1–2 days | Minimal. For runtime visibility, Relyance also needs access to the logs or gateway your apps call the models through. |
| Enterprise SaaS app | Salesforce, HubSpot, Greenhouse | Business system owners (sales ops, marketing ops, HR/People ops), IT admins | Medium — each app is technically simple, but every app has a different owner. The coordination is the work, not the connection. | About a day per app; 2–4 weeks end to end because owners need scheduling | Minimal. Some apps require an admin account or a higher licence tier to enable API access. |
| Data Subject Request | DSR portal, intake forms, request workflows | Privacy and legal, DPO, customer support, web team, IT | High — the most configuration-heavy piece. You decide how requesters are verified, which teams approve what, and how requests get fulfilled in each system. | 4–8 weeks, driven by decisions on verification, approvals, and fulfillment | A branded portal needs a DNS record (CNAME) for your subdomain and a TLS certificate, or an embed on your existing site. |
| Consent Management | Website and CMS, tag manager (Google Tag Manager, Adobe Launch), cookie banner, analytics and ad pixels, mobile app SDKs | Privacy and legal, marketing and web team, analytics or growth team, front-end engineering | Medium — adding the script is easy. The work is deciding what category each tracker belongs in and what gets blocked until someone consents. | 2–4 weeks; longer if you run many sites or regions with different rules | A JavaScript snippet on every page, and your tag manager must be set to respect the consent state before firing tags. Each domain and region needs its own configuration. Add the script host to any Content Security Policy allowlist. |
A common order is source code and cloud first, because they produce the map everything else hangs off. Data platforms come next, then AI and SaaS apps. DSR is usually last, since the workflows are easier to design once you can see where the data actually lives. Consent Management does not depend on the others and can run in parallel at any point.