Continuous Claude AI partnership and custom builds
After activation, Softblues stays on as your continuous Claude partner: new Projects, new MCP servers, new Skills, training refreshes, governance and a quarterly capability review that keeps your rollout in step with Anthropic and with your business. We run Claude across six of our own departments, so the builds come from people who use the platform daily.
Phase 3 continuous partnership at a glance
Softblues is a London-based AI practice and Registered Anthropic Partner Network member running ongoing partnerships after Claude Enterprise activation. The partnership covers five areas: new custom builds (Projects, MCP servers, Skills, sub-agents, computer use), training refreshes, governance and security maintenance, adoption support, and a quarterly capability review where leadership decides which new Anthropic features to adopt. The cadence runs weekly through quarterly, and the scope flexes between build-heavy, adoption-heavy and governance-heavy depending on what each quarter needs.
Leadership readout on new Anthropic capabilities, build catalogue updates and roadmap recommendations. The strategic anchor of the partnership.
Weekly, fortnightly, monthly, quarterly, annual. Predictable touchpoints from operational metrics to strategy refresh.
Projects, MCP servers, custom Skills, sub-agent orchestration, computer use agents, evaluation frameworks.
Custom builds, training refreshes, governance and security, adoption support and quarterly capability reviews.
Why you need an ongoing partner, not just a deployment
Companies that get long-term value from Claude treat it as a capability, not a one-off rollout. Three reasons an ongoing partner matters once Phase 2 is complete.
Claude is not a static product
Anthropic ships new capabilities every month: Skills, connectors, computer use, sub-agent patterns, models. A deployment current in January looks dated by July. Without someone tracking releases and matching them to your roadmap, your rollout falls behind the platform itself.
The build catalogue keeps growing
Phase 2 shipped the priority items. The rest sits in a backlog alongside new use cases that emerge as teams grow comfortable with Claude. Phase 3 ships them on a steady monthly cadence so the catalogue gets built, not shelved.
Adoption decays without a steady drumbeat
Champions leave. New cohorts join. Priorities shift. A 70% activation rate at month one drifts to 40% by month six without active maintenance. Weekly metrics, fortnightly champion calls, quarterly reviews. The cadence is what keeps adoption compounding.
What does ongoing Claude Enterprise support actually cover?
Five areas: custom builds, training refreshes, governance and security, adoption support, and a quarterly capability review. Every customer gets all five; the proportions flex by quarter.
Custom builds
New Claude Projects, MCP servers, Skills, sub-agents and computer use agents as your roadmap demands.
Training refreshes
New cohort onboarding, role-specific deep dives and refresher sessions for existing users.
Governance and security
Permission reviews, audit log analysis, policy updates and ongoing compliance posture work.
Adoption support
Weekly metrics, fortnightly champion calls, office hours and intervention when usage drops.
Quarterly capability review
Leadership readout on new Anthropic capabilities, build catalogue updates and roadmap recommendations.
What kinds of custom Claude capabilities does Softblues build for ongoing customers?
Six core build types: new Projects, MCP servers, custom Skills, sub-agent orchestration, computer use agents and evaluation frameworks. Most quarters ship two or three of these.
New Claude Projects
Per-function Projects bundling system prompt, knowledge base, Skills and connectors as adoption broadens.
MCP server development
Custom MCP servers in TypeScript or Python with auth, observability and rate-limit management built in.
Custom Skills
Skills that encode pricing rules, document templates and specialist workflows as reusable, governed assets.
Sub-agent orchestration
Multi-agent and multi-step workflows where Claude plans, delegates and verifies its own work end to end.
Computer use agents
Closed legacy systems with no API automated through Claude computer use, with safety and audit baked in.
Evaluation and tuning
Regression tests, quality gates and prompt tuning so each shipped capability stays accurate over time.
How does the cadence of the partnership work?
Weekly metrics, fortnightly champion calls, monthly delivery reviews, quarterly capability reviews with leadership and an annual strategy refresh. The rhythm is steady and predictable.
Weekly
Adoption metrics shared with the champion network. Office hours open to any user with a question.
Fortnightly
Champion network call across all functional champions. Issue triage, wins shared, blockers raised.
Monthly
Delivery review with engineering. What shipped, what is in flight, what is next. Scope check.
Quarterly
Leadership capability review. New Anthropic releases, roadmap updates, scope adjustments.
Annually
Strategy refresh that rebuilds the Phase 1 strategy against the past year and the year ahead.
How does the partnership shape itself to different customers?
Three common profiles: build-heavy in active rollout, adoption-heavy when usage is uneven, and governance-heavy in regulated environments. Most customers move between them across quarters.
Build-heavy
Active rollout or new departments adopting. Most capacity goes to new Projects, MCP servers and Skills. Typical in the first two quarters after activation.
Adoption-heavy
Usage is uneven across functions. Capacity shifts to champion enablement, office hours, training refreshes and targeted interventions where adoption has stalled.
Governance-heavy
Regulated environments or audit season. Focus on permission reviews, retention policies, audit log analysis, compliance documentation and security posture updates.
The partnership picks up where Phase 2: Deployment and Activation ends, with the build catalogue from Phase 1: Strategy and Roadmap as the starting reference.
External reference: Anthropic publishes its capability release notes that drive each quarterly review.
Where is the continuous partnership not the right fit?
If your deployment is finished, your champions are in place and your roadmap is stable, you may not need an ongoing partner. If you have an internal AI engineering team that can take it all on, we will hand over and step away. If you want a help-desk style support contract, that is not what we offer. We run ongoing partnerships where there is real, recurring engineering and adoption work. We do not run break-fix support, and we do not bill for unused capacity quarter after quarter. The relationship reviews itself every three months and either continues with a clear scope or pauses cleanly.
Common questions about the continuous partnership
How is the continuous partnership structured?
As a monthly engagement scoped to your needs. Softblues sets a fixed scope per month covering custom builds, training refreshes, governance reviews and adoption support. The scope is reviewed every quarter and adjusted based on what your business actually needs.
Do we need a Phase 3 partnership after deployment?
Not necessarily. Some companies activate at week 10 and run on their own with their internal champions. Softblues is happy to step away. Most of our customers stay because Anthropic ships new capabilities every month, the build catalogue evolves, and quarterly reviews keep the rollout moving forward.
What capabilities does Softblues actually build for ongoing customers?
New custom Projects as new functions adopt Claude. Additional MCP servers as new systems get integrated. Custom Skills for specialised workflows. Computer use agents for closed-system automation. Sub-agent orchestration for multi-step processes. The detail depends on your roadmap.
How often do we meet?
Monthly delivery review with engineering. Fortnightly champion network calls. Quarterly capability review with leadership where Softblues presents what new Claude features mean for your roadmap and what we recommend doing next.
Can the partnership flex up and down month to month?
Yes. We expect a baseline of governance and adoption support every month, with build capacity that flexes based on what your roadmap calls for. Some months are heavy on builds, some are light. Softblues sets expectations at quarterly reviews.
What happens if Anthropic ships something that changes our roadmap?
That is exactly what the quarterly capability review is for. Anthropic ships new model capabilities, new connector standards and new Skills frequently. Softblues assesses each release against your roadmap and recommends whether to bring it in, defer it or pass on it. You get to decide.
Three ways to start
Book a 20-minute call
A no-pitch conversation about where Claude could fit in your business. A Softblues team member will be on the call.
Book a callTake the Claude Readiness Score
Score your company's readiness for Claude Enterprise and get a plan back.
Start the assessmentGet the Buyer's Guide
A reference on platform choice, governance, costs, rollout timing and the questions to ask any vendor.
Download the guide