Back to News
News
AI can draw the blueprint. Who builds the workflow?
ChatGPT and Claude can help plan automated workflows, but dependable implementation needs connected systems, tested rules and clear ownership, which is where Reson8’s specialist platform approach comes in.
Share
Ask ChatGPT or Claude to design a workflow from enquiry to invoice and you can receive a surprisingly convincing plan.
It might describe how to capture a lead, update the CRM, score the opportunity, draft a proposal, prepare a contract and raise an invoice. It can suggest platforms, outline the connections and help write the code.
That's useful work. It can make the planning process considerably faster.
But a clear explanation of an automated business is still some distance from an automated business you can depend on.
For owners considering this approach, there are two decisions to understand: how the plan becomes a working system, and whether that system should connect specialist platforms or centre on an all-in-one product such as HighLevel.
What ChatGPT and Claude can help you design
AI assistants are useful thinking partners when you explain how your business operates.
They can help map the journey from first enquiry to payment, identify missing information and suggest where approvals belong. They can propose CRM fields, outline lead-scoring criteria and prepare test scenarios.
For an events business, that might mean distinguishing a general enquiry from a qualified opportunity with a date, budget and workable brief. For a consultancy, it could mean identifying which service a prospective client needs before assigning an adviser.
AI can also help challenge the sequence. Should a contract follow an approved proposal? Does an invoice require a completed signature? Who handles an enquiry when the usual salesperson is away?
These questions are the architecture: the rules that determine what can happen, when it can happen and who is responsible.
The quality of the answer depends on the information you supply. An assistant cannot infer your actual commercial policies reliably from a short prompt.
A chat response cannot deploy the system on its own
Saying that ChatGPT or Claude “cannot implement anything” would go too far. With authorised tools and an appropriate environment, AI assistants can help write, configure and test parts of a solution.
The important distinction is access and execution.
A normal chat response, without those connections, cannot authenticate your accounts, publish live automations or verify that a contract actually reached the customer. Generating instructions or code does not establish that they have been run successfully.
OpenAI's function-calling documentation describes a model requesting a tool action, with application-side code executing the function. Anthropic similarly distinguishes client tools, executed by the application, from tools run on its own infrastructure.[1][2]
For your workflow, a connector, application or authorised automation platform still needs to carry out the relevant action in HubSpot, DocuSign or Xero.
Even where AI helps perform the build, someone must establish the permissions, validate the business rules and take responsibility for operation.
That's why “AI can help design, build and test this” needs context. It describes assistance within a configured environment, rather than a promise that a prompt will deliver a production system.
What the working architecture looks like
For a workflow like the one below, Reson8's approach is to give each platform a defined job and connect the necessary steps.
This is an illustrative design, subject to checking the available actions, permissions and subscriptions in the selected accounts.
Stage | What happens | What must be controlled |
|---|---|---|
Receive enquiries | Capture website forms, emails and chatbot conversations; use AI to help extract requirements. | Preserve the original enquiry and flag uncertain or missing details. |
Update the CRM | Find or create the contact and company, then record the opportunity. | Define matching rules and prevent duplicate records. |
Assess the lead | Apply agreed criteria for fit, urgency, budget and potential value. | Record reasons; let scores guide the team rather than automatically reject prospects. |
Respond and assign | Acknowledge receipt, assign an owner and set a response deadline. | Separate routine acknowledgements from commitments requiring judgement. |
Prepare the proposal | Combine approved services, discovery notes and controlled pricing. | Approve bespoke scope, prices and the version being sent. |
Prepare the contract | Populate an approved template and send for signature at the agreed stage. | Use the correct customer, terms, document version and recipients. |
Create the invoice | Raise the deposit invoice after the required acceptance or signature. | Check the commercial trigger and prevent duplicate invoices. |
Follow up and report | Monitor proposals, signatures and payment status, then update records. | Stop reminders when circumstances change and distinguish sales value from money received. |
HubSpot could hold customer records and the sales pipeline. Airtable might hold structured operational information, while ClickUp manages delivery tasks. DocuSign handles signing, Xero holds accounting records and Google Drive stores approved supporting documents.
Zapier connects the required events and actions. Its published integrations include HubSpot with Xero, DocuSign connections and Airtable with ClickUp.[3][4][5] AI can sit within that workflow to interpret an enquiry or draft text.
Using every platform isn't the objective. If one tool already meets the requirement, adding another may create unnecessary work.
Why Reson8 often favours connected specialist platforms
For a business already using Xero for finance, ClickUp for projects or HubSpot for sales, connecting those tools can be a better starting point than replacing them.
Our preference for this kind of workflow rests on practical considerations.
Keep the functions your teams already rely on. Finance can continue working in its established accounting system. Sales and delivery can retain the tools suited to their responsibilities. That can reduce migration and retraining, although the integrations themselves still need careful design.
Make authority clear. The CRM owns the opportunity status, the signing platform owns the signature status and the accounting system owns invoice and payment records. Other tools reference those records rather than inventing competing versions.
Change individual components as requirements evolve. A modular design can allow a business to replace one tool while retaining others. It doesn't make switching effortless: fields, connections and historical records still require migration work.
Build around the actual process. A property business, events company and professional services firm may share a sales pipeline but have very different delivery requirements. A connected approach gives us room to accommodate those differences.
These are architectural judgements, not proof that one vendor is always superior.
Where HighLevel fits
HighLevel is a software platform; Reson8 provides design and implementation services. Comparing the two directly can obscure the real choice: the architecture that best suits the business.
HighLevel presents an integrated offering covering CRM, forms, websites, conversations, workflows, proposals and invoicing.[6] Bringing related functions into one environment may suit businesses seeking a consolidated sales and marketing operation.
It also connects to Zapier through LeadConnector, and its documentation describes an AI workflow builder.[7][8] It would therefore be inaccurate to suggest that HighLevel cannot automate or connect external tools.
For the cross-department workflow described here, Reson8's case for specialist platforms is the ability to preserve suitable existing systems and give each function a clear home. A business with established accounting and delivery processes may gain little from migrating them simply to consolidate its software.
HighLevel may still be a sensible choice when its included capabilities match the requirements and consolidation reduces overall complexity.
There are costs on both sides. Multiple tools mean multiple subscriptions, connections and potential points of failure. An integrated suite can concentrate dependence on one supplier. Compare the complete requirements, ongoing costs and exit arrangements before deciding.
Zapier is also optional where a suitable native integration already does the job reliably. The choice should follow the workflow.
The implementation is where reliability is earned
Consider a signed contract arriving twice because an event is retried. Should the business issue two deposit invoices? Clearly not.
The system needs a way to recognise that the invoice already exists. It also needs to handle a revised proposal, a declined contract, an expired connection and an incomplete enquiry.
Those cases are part of the build.
Prices and payment schedules should come from approved records and calculations. AI can draft the explanation around them. Bespoke commercial commitments should pass through the agreed approval process.
Launch should begin with controlled examples and reviewed drafts. Routine cases can then be automated further once the team has evidence that the workflow behaves correctly.
Someone also needs to monitor failures, maintain access and update the process when the business changes.
At Reson8, that is the work between a convincing blueprint and a useful everyday system. We review the process, choose suitable platforms, connect them and test the complete journey.
Talk to Reson8 about turning your workflow ideas into a connected operation that gives your team more time and helps fewer opportunities slip through the gaps.
Sources
Official documentation checked on 15 October 2026. Platform suitability and the recommended architecture are Reson8's editorial assessment; specific actions and subscription requirements must be verified before implementation.
Share
Back to News