What You Need to Know
Integration, in this certification, asks where in an existing process a person brings work to Claude, and whether the process keeps its shape. It never means a technical connection between systems. To augment is to hand Claude one slow step. To redesign is to change the process itself. To escalate is to pass work beyond the Associate's role to someone else.
Take a dry cleaner. Every Monday its manager writes up last week's damage claims. Customers who leave garments uncollected get a reminder letter.
Augment: One Step, Done Better
When Claude takes over the slow step, everything else keeps its shape. The write-up takes ten minutes instead of an hour. The manager still owns it, and nobody else's job changes. So augmentation is cheap, reversible and easy to explain to colleagues. If it fails, the manager writes it by hand again. It is the right choice far more often than the ambitious alternative. Most successful integrations for an Associate are one step, done better.
Redesign: Pilot First
In a redesign the process changes because the constraint that gave it its shape has gone. Personalising a reminder letter used to cost twenty minutes, so the dry cleaner sent one generic letter. At two minutes, a tiered approach nobody would have proposed becomes possible.
Redesign is rarer and bigger, so it gets a pilot before a rollout. Anthropic's enterprise deployment guidance stages adoption this way: a pilot group monitored for feedback over a period of weeks before wider expansion, with champions identified per team. The dry cleaner tries tiered letters on one batch first. An all-at-once rollout measures nothing, with nothing left to compare against, and risks everything at once.
The Human Gate
Find the step where the decision happens, and keep a person on it. Integrations fail when the checking step is removed. Checking has no visible output, so it looks like waste to anyone redesigning from a diagram. The dry cleaner's manager signs each letter before it goes out. The gate also makes augmentation safe to adopt quickly. The person who always signed the letters still signs them, so the judgement never left the process.
The Associate's Boundary
Some requests leave this role because they need a technical implementation: systems talking to each other, data moving on its own, or a process built to run with nobody present. Suppose the owner wants reminders sent overnight, straight from the till system. That means connecting the till system to whatever sends the letters, which is technical work. The professional move is to escalate to a Claude Developer or Architect with a clear written description of what the process needs. Describing the need well is itself the Associate's contribution. Escalation is no failure. It is knowing the shape of your own job.
Connectors let Claude retrieve data and take actions in connected services. They operate inside a conversation, when a person's requests call for them. The Google Workspace documentation states that Claude only accesses your data when you explicitly ask a question or request an action requiring it. Within the features this exam covers, such as chat, Projects, research, artifacts and connectors, work moves when a person asks for it.
Current state (checked 14 October 2026): Claude Cowork, which the exam guide does not name, has scheduled tasks. Anthropic's Cowork course says a task that only uses your connectors can run remotely at its set time. On the exam, a request for work that runs by itself is still escalated, because the exam treats it as a technical implementation beyond this role.
Where Recurring Work Lives
Recurring work belongs in a Project, not in one ever-lengthening conversation. A Project is a self-contained workspace with its own chat histories and knowledge. When a chat approaches its length limit, the Help Center advises starting a new conversation or using a Project. The Monday write-up goes there. Where a plan supports sharing, a Project gives a team one shared environment for its documents and instructions. Chats inside a Project are not shared by default. A member sees the Project's contents, knowledge and instructions. Each person's own conversations stay private unless they choose to share one. Deciding the work belongs in a Project is this objective. Configuring the Project and maintaining its instructions and knowledge is Domain 5's.
In the exam
The meaning of integration answers many stems. The distractors are usually integrations in the technical sense. The right answer is an integration in this lesson's sense. Learn the stem triggers that signal a technical implementation, and so an escalation: automatically, without anyone having to, straight from the system, overnight.
The automation leap
Wiring Claude into a system, or setting it up to run unattended, is this domain's signature distractor and is never the right answer. Escalate because the request is a technical implementation, beyond this role and the features this exam covers. Never rest the escalation on a claim that no Claude product could run the work. Cowork's scheduled tasks, not named in the exam guide, can run some tasks at a set time.
Exam Traps
Improvising DIY automation instead of escalating
Systems wired together, auto-publishing and unattended pipelines are technical implementations, beyond the Associate's role and the features this exam covers. Improvised versions fail silently with nobody present to catch them. The professional move is a clear escalation brief to a Claude Developer or Architect.
Redesigning the whole process at once
The boil-the-ocean rollout means that if it fails, everything fails together and nothing was learned. Redesign earns a pilot on one strand first, with a measurement, exactly as Anthropic's own deployment guidance stages adoption.
Optimising away the checking step
The human gate produces no visible output, which makes it look like waste on a process diagram. It is where the process's judgement lives; remove it and the first uncaught error becomes the integration's reputation.
Practice Scenario
A local newspaper's events page is compiled every week by an editorial assistant from event details that venues email in. A colleague proposes that Claude should instead gather the listings from the venues' websites overnight and publish the finished page without anyone touching it. The assistant still wants to save time on this week's page, with a person reading every listing before it runs. How should the assistant respond?
Build Exercise
Map One Process You Own
What you'll learn
- How to classify each step of a process: augment, redesign, escalate or human-only
- How to write a usable escalation brief for a Developer or Architect
- How to pick the augmentation with the best ratio of time saved to risk
- How to measure an integration honestly over two weeks
- Pick a recurring process you could describe to a new colleague in five minutes, and write out its steps end to end
Why: Integration decisions are made per step; a process you cannot enumerate cannot be integrated deliberately.
You should see: A numbered list of steps from trigger to finished output, including the informal ones.
- Mark each step: augment (a person brings this step to Claude today), redesign (this step only exists because something used to be expensive), escalate (this needs a technical implementation, such as systems talking to each other or a process running with nobody present), or human-only (this is where the decision happens)
Why: The four labels are the objective’s whole decision framework applied to real work.
You should see: Every step carrying exactly one label.
- For every escalate step, write the two-sentence brief you would hand a Claude Developer: what the process needs, and what must stay human
Why: Describing the need clearly is the Associate’s contribution to work beyond the toolkit, and writing it tests whether you understand the process.
You should see: A brief per escalate step that a Developer could act on without a meeting.
- Pick the single augmentation with the best ratio of time saved to risk and run it for two weeks, keeping the human gate exactly where it was
Why: One step, done better, measured, is how integrations succeed; the pilot discipline applies even at this scale.
You should see: Two weeks of the process running with one step augmented and nothing else changed.
- Measure it honestly (the before, the after, how you know) and note anything the redesign question should revisit now that the step is cheap
Why: The honest measurement feeds objective 4.5, and the cheap step may have removed a constraint the process was shaped around.
You should see: A three-line record: before, after, evidence, plus any redesign candidate worth piloting.
Sources
- What are projects? — Anthropic
- Manage project visibility and sharing — Anthropic
- Use connectors to extend Claude's capabilities — Anthropic
- Use Google Workspace connectors — Anthropic
- How do usage and length limits work? — Anthropic
- Claude Enterprise Administrator Guide — Anthropic
- Introduction to Claude Cowork: What Claude Cowork can do for you — Anthropic (current-state note only: the exam guide does not name Cowork)
- Claude Certified Associate – Foundations Exam Guide — Anthropic (exam guide; claims about professional practice are our judgement, not Anthropic guidance)