Automation and AI support
Less chasing information. Fewer repeat questions. More time with clients, making decisions and moving work forward.
We help businesses connect their tools, automate repeat jobs and put AI to work inside the team's day. That could mean better HubSpot follow-up, an assistant people can ask for help in chat, or a faster way to turn meeting notes into the next steps.
Start with the job that keeps eating your time. You don't need a technical brief.
When everything comes back to you
Someone needs the background on a client. Someone else is unsure how a system works. A meeting's finished, but the notes still need turning into an email, tasks and a CRM update.
Each job looks small. Together, they fill the gaps in everyone's day. The more experienced people get interrupted, while the rest of the team waits for an answer or pieces it together themselves.
An assistant with access to the right information can help people check an approach, find the source and prepare the work. Automation can handle the predictable steps that shouldn't need another person to remember them.
The opportunity is to let your team spend more time reviewing, deciding and managing the work, with less of the preparation landing on the same few people.
Inside CLCK
We've put an AI assistant inside our own business. We call it Arlo, and it's built around Hermes Agent, the open-source framework from Nous Research.
In Slack, anyone on our team can mention Arlo in a conversation and ask for help. It can work from the information and tools we've given it access to, rather than relying on someone to paste the whole backstory into a new chat.
That includes drafting client responses, preparing documents, checking technical questions and helping with approved work in HubSpot. The team can see the request and the answer in the same thread, within the access boundaries for that work.
One of our team members asked Arlo to check a website's HubSpot tracking and domain connection. It separated the two issues: the domain was connected, but tracking wasn't detected in the homepage test.
That was a scoped check, rather than a claim to have tested every page or fixed the site. It gave the team a better starting point for the next decision without waiting for a technical colleague to investigate first.
We're spending more time checking the work and managing the next step. People still own the client relationship, the judgement and the approval. Arlo helps with more of the work in between.
Read what we've learned using an assistant inside CLCKLess explaining from scratch
If you've used AI for work, you've probably spent time explaining the same things again. Who the client is. How you write. What belongs in a meeting brief. What it must check before calling something finished.
In a configured assistant, selected context and agreed instructions can carry forward. Hermes has a “soul” file that helps shape behaviour, memory for selected information, and reusable instructions called skills.
A skill can describe where to look, which steps to follow, what the finished work should contain and when to ask for approval. You can start with an existing procedure or work through a task together, then ask for the agreed approach to be written up for review.
Check the agreed project records. Summarise what was completed. List what needs a decision. Flag missing information. Draft the update in our usual format, ready for someone to review.
The next person asking for that job has a shared starting point, rather than their own collection of prompts.
When you find a better way, you can review and approve a change to the instructions. That doesn't retrain the underlying model or guarantee perfect results. It means the effort you put into explaining the work can help beyond one conversation.
Where it can help
These are starting points to assess against your systems and permissions, rather than a promise that every connection is ready out of the box.
Instead of opening the CRM, searching old emails and hunting for call notes, ask for a briefing from the agreed sources. What does the prospect need? What did you promise last time? What still needs an answer? Your team gets a starting point they can check before the call.
Use the transcript to prepare a summary, actions with owners and a follow-up email for review. The instructions can tell the assistant to flag a missing deadline rather than invent one. You spend your time checking the decisions and the tone, instead of starting with an empty document.
Find missing fields, stale deals and opportunities without a next step. Prepare a clean-up list for the team, or use agreed automation to route work and create reminders. Changes to records stay within the access and approval rules you've set.
Pull together the agreed project notes, completed work and outstanding questions, then prepare an update in your usual format. The person managing the client checks it and approves the message. Less Friday afternoon spent reconstructing what happened all week.
Turn a rough idea into a brief, a draft and supporting content. Keep the sources, brand instructions and review points with the work. That's how we approach our own marketing: people choose the direction and approve the output, with Arlo helping prepare the pieces.
Where reporting connections are available, scheduled checks can bring a summary into the team's workspace. Review changes in spend, enquiries or pipeline without manually opening every dashboard. Recommendations come back for a decision; they don't become permission to change a campaign.
A working example from our own marketing
Our marketing work often starts with a note, an idea or a question about what deserves attention next. Arlo can help inspect the content queue and existing pages, gather source material and prepare a draft.
The review stays with us. We check the argument, the facts and whether it sounds like something we'd actually say. After approval, page preparation can include the layout, internal links, metadata and mobile checks.
For larger jobs, we keep the main conversation focused on the goal and split detailed tasks into smaller pieces. Each piece needs a result that can be checked before we move on.
That leaves less manual coordination between an idea, a document, a web page and the decision to publish it. It also makes the unfinished work easier to see.
This is our own operating experience, not a claim that AI runs our marketing unattended. We still review the output, catch mistakes and approve what goes live.
See how we use Arlo for CLCK's marketingChoose the right tool for the job
If the same event should trigger the same action every time, straightforward automation is often the better answer. Assign an enquiry to the right person. Create a task when a deal reaches a stage. Start an approved nurture sequence.
AI is worth considering when the task involves reading, gathering context, preparing a draft or working through a less predictable question. It needs good sources and a way for someone to check the result.
The two can work together. A workflow might flag a deal without a next step; an assistant could then prepare a summary of the relationship for the salesperson to review.
And sometimes neither is the first fix. If nobody agrees who owns a lead, automating the handover can make the confusion happen faster. We sort out the ownership or data problem before adding more machinery.
Talk through the work you want to simplifyClient experience
Our client work in CRM and automation shows what happens when the tools fit the way people work. These are HubSpot and process-improvement examples, separate from our use of AI inside CLCK.
Super Property had a collection of systems that needed a more connected approach. We implemented HubSpot and automated key sales and operational processes, including text message integration.
For the team, the benefit was less admin around the sales process and improvements that reached across sales, marketing and operations.
“Damien's expertise gave us immediate confidence. He transformed our 'duct taped' systems with a strategic HubSpot implementation that automated our sales process and saved our team a huge amount of time. We've seen significant improvements across sales, marketing, and operations as a result.”
The parish needed a central place for information and a system that reflected how its team worked. CLCK helped with the HubSpot structure, training and rollout, replacing scattered spreadsheets and paper with a shared source of truth.
The starting point was understanding the people and the process, so the new setup made sense to the team using it.
“Damien was incredibly patient, taking the time to fully understand our needs before recommending the right approach. His expertise made the entire transition smooth.”
We built automated lead nurture and improved data visibility for a manufacturer whose revenue grew from roughly $8 million in 2015 to $40 million in 2020.
That happened over five years, with CLCK's work supporting the wider business effort. It wasn't an AI result or growth attributable to automation alone. The lesson is that consistent follow-up and better information can contribute well beyond the first campaign.
Explore more client resultsAccess and judgement stay with people
Connecting an assistant to the business doesn't mean opening every system to it. We start with the information and actions needed for the agreed jobs.
Written instructions matter, but they aren't a substitute for tool permissions and human checks. We discuss the controls available in your systems, how the chosen AI providers handle data and what information shouldn't be used.
Our own rule is simple: preparation and permission to publish or send are different things. A convincing draft still needs the right person to approve it.
Getting started
We talk through where work slows down, what gets repeated and who keeps getting interrupted. Choose a task with enough frequency and a result you can check.
We look at the tools you already use, the information the job needs and the permissions available. Decide what stays manual and who reviews the result.
Set up the agreed workflow or assistant connections, instructions and checks. Test it on representative work, including what should happen when information is missing.
Walk people through how to ask for help, check the output and report a problem. Agree who looks after the setup, connections and instructions as the business changes.
Compare the time spent before and after, including review and corrections. Look for fewer missed handovers and follow-ups, and include running costs. Expand when the first job earns its place.
A ten-minute draft that takes an hour to fix hasn't saved you time. We measure the finished work, rather than how quickly AI produces an answer.
Explore the part that matters to you
Before you decide
A chatbot usually answers what you type into it. An AI agent can also work with connected tools, documents and business information to prepare or carry out agreed tasks. What it can do depends on the connections, permissions and instructions in place. We start with the job your team needs help with, then choose the right approach.
More enquiries won't solve a follow-up problem. Automation can route leads, create reminders and keep ownership visible. AI can help with account research, meeting preparation and follow-up drafts. Together, they can help your team spend less time gathering information and more time having the conversations that matter.
Yes. HubSpot is often where contact records, deals, tasks, workflows and reporting come together. We can help with HubSpot automation and assess which AI connections fit the work. The exact actions available depend on your HubSpot subscription, supported connections and agreed permissions.
An assistant can work inside a supported team chat app. We use Arlo in Slack, where team members can mention it and ask for help in a shared thread. People still need a walkthrough of what it's for, what it can access and how to check its work. We confirm your chat platform and connection requirements before recommending a setup.
Selected business context and approved working instructions can carry forward between tasks. In Hermes, memory and reusable skills help with that. A skill can describe the sources to check, the steps to follow and the output to prepare. This isn't automatic perfect recall or retraining the underlying AI model. We agree what should be retained, keep private context within its intended scope and review changes to shared instructions.
You agree the systems, data and tasks it can access. We define where it can inspect or prepare work, where human approval is needed and who can give that approval. Instructions are paired with the access controls available in the connected tools. Sensitive information, external messages and live changes need particular care; an AI assistant should never have unrestricted access simply because a connection is possible.
No. We usually start with an existing job that causes delays or repeat admin. Sometimes a HubSpot workflow is enough. Sometimes the underlying data or process needs attention first. We look at what you already use and test the smallest change that can make the job easier before expanding.
People need to know how to use it, and someone needs to look after it. We discuss training, hosting where needed, software updates, connection issues and changes to working instructions as part of the scope. We also explain the relevant model-provider and software costs. The ongoing arrangement depends on what we build and what your team wants to manage internally.
Choose a job you can compare before and after. Look at preparation time, the effort needed to check the result, missed follow-ups and how often work has to be redone. Include running and maintenance costs. We don't promise a universal number of hours saved: the test is whether the finished workflow makes life easier for your team.
Let's start with your working week
The meeting prep? The repeated questions? The follow-up that only happens when someone remembers?
Tell us where the time goes. We'll talk through whether a workflow, an AI assistant or a change to your existing setup could help, and what a realistic first step looks like.
Book a strategy sessionNo technical presentation to prepare. No hard sell.