How We Implement Claude Cowork — the Tinkso Approach

This page is for buyers who’ve read the pitch and want the detail. Here is how we actually run a Claude Cowork implementation. If you want the short version first, the homepage has it.

Observe → Co-design → Ship

Every Claude Cowork deployment we run follows the same three-beat motion. It’s the single thing that separates deployments that stick from deployments that quietly die.

Observe

Before we build a single connector or skill, we shadow the people who’ll use them. Half-day observation sessions per role, across seniority levels, watching the real work happen. Shadowing surfaces every undocumented shortcut and hidden decision point interviews miss.

Co-design

Business users co-build skill specs with Tinkso engineers in real time. The user who’ll use the skill every day sits in the spec session, pushes back on defaults, and signs off on inputs and outputs. By launch, they’re already internal champions.

Ship

Three one-week production sprints — Foundation, Connect, Adopt — each ending in a visible deliverable. Plus the pre-engagement Discovery phase where Observe and Co-design live.

The four-phase engagement

Sprint 0 — Discovery (3–5 days)

Outcome: A workflow map you’d want even if you never deployed Claude.

• Shadow sessions with 2–3 users per function (half-day each)
• Decision-point interviews capturing the judgment calls nobody wrote down
• Pain/gain mapping per role
• Skill backlog prioritization derived from the map
• Co-design workshops with power users for top-priority skills

Deliverables: Workflow Map (one-pager per function), prioritized skill backlog, recommended MCP connector shortlist, fixed-fee proposal.

Sprint 1 — Foundation (week 2)

Outcome: Claude Cowork is set up properly and your pilot group is actively using it.

• Claude Enterprise workspace provisioning and org configuration
• SSO with your identity provider (Okta, Entra, Google, Ping)
• SCIM user lifecycle + group-based role mapping
• Admin policies: retention, residency, audit
• Baseline telemetry dashboard
• Pilot group (10–20 power users) activated

Sprint 2 — Connect (week 3)

Outcome: Claude Cowork is doing real work against your systems.

• Custom MCP connectors to your highest-value systems
• Packaged skills built from the co-designed specs
• Integration testing against real past cases
• Internal champion program activation — 2–3 per function

Sprint 3 — Adopt (week 4)

Outcome: Team-wide rollout, trained, measured, handed over.

• Role-specific training sessions per function (90 min each)
• Full-team enablement
• Office hours schedule for 60 days post-launch
• Day-30/60/90 adoption review cadence booked
• Handover call with your designated admin

What’s included at every tier

Standard — starting in the low five figures (fixed fee)

For clean environments, 2 core systems to connect, a single-digit number of functions to train. Includes full Sprint 0 Discovery, Foundation, 2 MCP connectors, 5 skills, training for up to 3 functions, 60 days post-launch support.

Complex — mid-five to low-six figures (fixed fee)

Extended Discovery across 4–5 functions, 4–6 MCP connectors, 8–15 skills, training across 4–5 functions, change management support.

Retainer — monthly, starting in the low four figures (post-launch)

Defined monthly connector + skill build capacity, weekly office hours, monthly adoption review, priority response SLA, quarterly strategic review. Pricing philosophy →

What we don’t do

• We don’t do Claude Code. Developer tooling is a different product with a different buyer.
• We don’t sell slide decks. Every engagement ends with shipped work.
• We don’t skip Discovery, even on rescue engagements.
• We don’t bill hourly. All work is fixed-fee after scoping.
• We don’t do sub-25-seat deployments. The engagement economics don’t work below that headcount, even if the Team plan technically starts at 5 seats.
• We don’t take 10,000+ seat enterprises. Anthropic’s own team will serve you better.
• We don’t do FedRAMP work today. HIPAA-ready deployments on Enterprise we can support.

How we price

Fixed fee, always

We scope, we price, we sign. After signature, the price doesn’t move. If scope needs to expand mid-engagement, that’s a separate scoped addendum — never a time-and-materials surprise.

What drives the price

• Discovery scope — number of functions to shadow
• System count — each MCP connector is a line item
• Skill count
• Auth/IdP complexity
• Data hygiene — systems may need cleanup before Claude can use them

Our principles (and guarantees)

• We listen before we build. Discovery isn’t optional.
• We ship in sprints, not in quarters. Four weeks end-to-end or we propose a different engagement.
• Your team does the final review. No work ships without your sign-off.
• Code and configs are yours. Every connector, skill, and config is handed over.
• We tell you when we’re wrong.

Roles — yours and ours

Risk, security, and compliance

We’ve shipped deployments through Big Four security reviews, regulated financial services IR teams, and AmLaw 100 legal ops. What passes:

• Auth: OAuth 2.1 with per-user delegation; no shared service accounts
• Data handling: MCP servers in your environment; no prompts or outputs stored by Tinkso
• Audit trail: every connector tool call logged — feeds your SIEM
• SOC 2 Type II: Tinkso is SOC 2 Type II audited
• DPA: standard DPA signed with every engagement; GDPR-aligned

After launch — the 60-day window and beyond

Days 0–60 (included)

• Weekly office hours
• Monthly skill showcase sessions
• Day 30 / 60 adoption review calls
• Direct Slack Connect or email support channel

Day 60 — decision point

1. Monthly retainer — ongoing connector/skill development
2. Quarterly tune-up — scheduled half-sprint every quarter
3. Full handover — you own everything

About 70% of clients go retainer after 60 days. Most stay for 12+ months.

What we need from you to start

• Signed SOW (standard MSA + SOW; we provide templates)
• 2–3 named users per function willing to be shadowed
• Admin access to your Claude Enterprise workspace
• IdP admin contact for SSO/SCIM configuration
• 1 executive sponsor for weekly 30-min checkpoints
• System access credentials for the systems you want connected

Most engagements kick off within 5 business days of signed SOW.

FAQ