AI Care Plans: keep your systems working after launch
Your AI setup worked the day it launched. That was the easy part.
Agents, automations, and integrations don’t sit still. Models get retired, prompts drift away from the business, connections quietly stop firing, and the whole thing keeps looking fine from the outside while doing less and less of what you built it for.
Clever Care for AI is the plan that keeps someone watching.
System monitoring and failure alerts
We watch every automation, agent, and integration in your setup. Failed runs, broken webhooks, expired tokens, revoked connections, and scenarios that quietly stopped firing. Most AI systems don’t break loudly. They just stop, and nobody notices until a client asks why they never heard back.
Model and platform version management
The companies behind these tools retire models on a published schedule. Anthropic commits to at least 60 days’ notice, OpenAI runs a similar window, and once a model is retired, every request to it fails. That notice goes to whoever’s email address is on the account, which is often a developer who’s moved on. We track the notices, test the replacement model against your real inputs, and move you across before the cutover rather than after it.
Monthly output review
We sample the actual work your system is producing and check it against what it’s meant to be doing. This is the closest thing AI has to the regression testing we run on websites, and it’s the inclusion clients can’t realistically do for themselves. Quality slides slowly enough that the people using it every day stop seeing it.
Context and brief maintenance
Your prices change. Your services change. Your team changes. The documents your AI works from need to change too, or it keeps confidently answering from last year. We keep the briefs, knowledge bases, and context documents current so the output tracks the business.
Access, keys & spend
Quarterly systems review
Once a quarter we look at the whole setup. What’s earning its keep, what should be switched off, and what the business needs now that it didn’t need when this was built. You get an actionable overview, not a dashboard.
Trust what your systems do when you're not watching
You built these systems so you could stop doing the work yourself. That only holds if someone is checking that the work is still being done properly. We keep your automations running, your models current, and your output honest, so you can leave the thing alone and get on with the parts of the business that need you.
How we look after your AI systems
Rocky foundations and lack of maintenance are fixable problems that can turn AI failure into tools that actually meet the promise they propose and have a positive impact on your business.
How it works
Step 1: AI Health Check
Any setup we didn’t build, or haven’t managed for a month or more, starts with a Health Check. We map what’s actually running, where it runs, who owns the keys, what it costs, and what’s already broken. Most people are surprised by at least two of those. If remediation is needed to bring things back to a working baseline, we’ll quote that separately before cover starts.Any setup we didn’t build, or haven’t managed for a month or more, starts with a Health Check. We map what’s actually running, where it runs, who owns the keys, what it costs, and what’s already broken. Most people are surprised by at least two of those. If remediation is needed to bring things back to a working baseline, we’ll quote that separately before cover starts.
Step 2: Remediation
We fix what the check turned up. Failed scenarios, deprecated models still in use, connectors holding more access than they need, and workflows nobody wrote down.
Step 3: Clever Care cover begins
We’ll consolidate your licenses where it makes sense. We’ll also outline the tasks we’ll handle for you and write the documentation so your setup isn’t living in one person’s head. Any add-ons you want can be included at this point.
Step 4: Review & improve
We monitor, review the output monthly, and sit down with you quarterly. When something needs building or changing, you can book that as you go rather than paying for hours you might not use.
Nicole and team have great ideas and we love working with them!
Model retirements are scheduled & requests to a retired model fail. Anthropic gives at least 60 days' notice before retiring a publicly released model. In the last year alone, Claude 3.7 Sonnet, Claude 3.5 Haiku, Claude Sonnet 4, and Claude Opus 4.1 have all been retired. OpenAI shut down a batch of older models in July '26 and has flagged more for Dec '26 and Jan '27. Migration isn't optional; it's just a question of whether you do it before or after your system stops working. Sources: Anthropic model deprecations, OpenAI deprecations.
AI projects fail at more than twice the rate of other technology projects. RAND researchers James Ryseff and Anu Narayanan put the failure rate above 80%, and found that the models themselves were rarely the problem. It was the data and the systems underneath them decaying. Source: RAND Corporation.
Broken systems don't tell you they're broken. One client spent months wondering why her products weren't selling online. Her purchase options had become disconnected from a payment processor. There was no error and no complaint, just an absence she'd been reading as a slow market. The only reason anyone finds a failure like that is because someone is looking for it.
Get your AI systems covered
No complicated tiers. One plan that covers your setup, plus a fee for each live workflow, because looking after eight of them takes more than looking after two.
Clever Care for AI
Covers your account, run monitoring, spend reporting, context maintenance, monthly output review, and the quarterly systems review.
AUD$450 + GST (~US$325)/month (includes one live workflow)
Additional live workflows: AUD$120+ GST(~US$86) each per month
AI Health Check
Required before cover starts on any setup we didn’t build. Includes the full inventory and a written report you keep, whether you go ahead with a plan or not.
AUD$750/month + GST (~US$500)
Book a conversation
Add-ons
Bulk Hourly Packs
On-demand access to web, systems, and growth help. Valid for 12 months, timed in 15-minute increments, for when you need something sorted quickly. 10h: AU$1800 · 20h: AU$3400
Agent tune-up
A focused pass on a workflow that’s underperforming. We rework the brief, retest against real inputs, and hand back something that does the job properly again.
Model migration
When a model you rely on is retired, or when a newer one would do the job better or cheaper, we handle the swap and the testing as a defined piece of work.
Team enablement session
A working session with your team on using the systems properly, spotting when something’s gone wrong, and knowing what to escalate.
Usage and cost review
A deep look at what your AI stack costs and where the spend is going. We right-size plans, cut what isn’t used, and route work to cheaper models where the quality holds.
Build something new
Need a new workflow rather than care for an existing one? That’s a Clever Sprint.
Frequently asked questions
What actually breaks in an AI or automated system if nobody touches it?
More than people expect, and almost none of it is obvious. Models get retired and requests start failing. Integrations lose their authorization when someone changes a password or leaves the company. Prompts and briefs stay accurate while the business moves on around them, so the output slowly stops matching reality. Token usage costs creep as volume grows. And connectors keep whatever access they were granted on day one, long after they need it. The common thread is that the system keeps looking like it’s working.
Can't I just check this myself?
Absolutely, and here’s what that involves:
- Watch every automation for failed runs, not just the ones that alert you.
- Subscribe to deprecation notices from every model provider you use and add the retirement dates to your calendar.
- Test replacement models against your real inputs before each cutover.
- Sample your system’s output monthly against what it should be producing.
- Keep every brief and context document current as your services and prices change.
- Rotate API keys and audit connector permissions.
- Track your usage spend and check it against what you’re getting.
Most people do this properly for about two months.
Does this include building new agents or automations?
No, and that’s deliberate. We don’t include build time in the base plan because we don’t want to charge you for hours you may not need. New workflows are a Clever Sprint, smaller changes come out of a Bulk Hourly Pack, and anything urgent can be booked on demand.
What if you didn't build my setup?
No problem, that’s the case for most of the setups we see. Everything starts with an AI Health Check, so we know exactly what we’re taking on. If it needs work to get to a stable baseline, we’ll quote that first.
Who pays for the API and platform costs?
You do, on your own accounts, and that’s the right way around. Your data and your workflows shouldn’t sit inside our billing relationship. Where we hold agency licenses for tools you use, we pass the savings on the same way we do for our WordPress clients. We report on your spend monthly so there are no surprises.
Do you have access to our data, and where does it go?
We have the access we need to do the job, and no more. As part of onboarding we document exactly which systems we can reach and what each connector can see, and we review those permissions quarterly. We don’t move your data into our own tools, we don’t train anything on it, and if you end the plan we hand back the documentation and remove our access.
What happens if a model gets retired?
We see the notice, test the recommended replacement against your real inputs, and migrate you before the retirement date. If the replacement behaves differently enough to need the brief rewritten, we’ll tell you and quote that as a migration. Your system won’t fail because a date passed.
What if the person who built this has left?
This is very common, and it’s one of the better reasons to get a Health Check done. We reconstruct what’s running, document it properly, and take ownership of the accounts and keys so the knowledge lives somewhere other than a former employee’s laptop.
Can you look after automations that don't involve AI?
Yes. Most real setups are a mix of integrations, automations, and a few AI steps, and drawing a line down the middle of that would be unworkable for everyone. Clever Care for AI covers your automation layer, AI included. Some of our clients have us managing automation only, with a future-ready approach to watching where AI should slot into workflows and when to flick the switch.
Do I need to be a Studio Clvr client?
No. Unlike our WordPress care plans, this one is open to systems we didn’t build, as long as we’ve done a Health Check first.