How I Ship with AI Every Day: Tools, Workflows and Real Use Cases
By Umar Din · Prime AI Solutions · Published 4 March 2026
Most writing about AI automation describes what is theoretically possible. This post describes what I actually do. The tools I open every morning, the workflows I build for clients, the industries I build them in, and what I have learned about what works and what does not.
If you are an operations director, finance manager, HR lead, or business owner trying to understand what an AI automation consultant actually does, this is for you.
What I Actually Do Day-to-Day
The work breaks into three phases for every client engagement.
First, workflow mapping. Before touching any tool, I spend time understanding how the business actually operates: what tasks happen daily, who does them, how long they take, what data moves between systems, and where the bottlenecks are. This is where most of the value is generated, not in the build phase.
Second, tool selection and build. I select the right automation platform for the job based on the client's technical capability, data requirements, and existing tool stack. Then I build, test, and integrate the automation. This part is visible. The next part matters more.
Third, handover and training. An automation that the client cannot understand or extend is a liability. Every engagement ends with training the team to manage, monitor, and expand the automation themselves. The goal is to make myself unnecessary.
The Tool Stack: What I Use and Why
I have hands-on experience with all of the tools below. Here is what each one is, what I use it for, and a realistic use case from client work.
GPT-5 / ChatGPT Canvas (OpenAI)
What it is: OpenAI's most capable model, with ChatGPT Canvas providing a collaborative document workspace for iterative writing and analysis inside the ChatGPT interface.
What I use it for: I use GPT-5 as the default reasoning layer for client-facing content, financial analysis tasks, and any scenario requiring strong general-purpose intelligence. ChatGPT Canvas is particularly useful for drafting and refining documents with stakeholders who are not technical, as it feels like a word processor rather than a chat interface.
Example use case
For a CFO preparing a board-level AI business case, I used ChatGPT Canvas to co-draft the document in real time, iterating on assumptions and language across multiple sessions. The final output was used directly in the board pack.
Claude Sonnet 4.6 (Anthropic)
What it is: Anthropic's frontier model, exceptional at long-document reasoning, structured analysis, and nuanced writing. Claude Sonnet 4.6 is the model I use most in day-to-day consulting work.
What I use it for: I use Claude for anything requiring sustained reasoning across large documents, careful analysis of financial or legal text, or where output quality consistency matters. Three Anthropic tools extend Claude's capability significantly: Claude Browser Extension for in-browser tasks, Claude Code for software and automation build work, and Claude CoWork for setting agents to work on multi-step tasks autonomously.
Example use case
Using Claude CoWork, I set agents to audit a client's existing process documentation overnight: extracting automation opportunities, categorising by complexity, and drafting a prioritised opportunity matrix. Work that previously took two days of analysis was ready by morning.
NotebookLM (Google)
What it is: Google's AI research and synthesis tool that allows you to upload documents, reports, and data sources and then query them, summarise them, or generate audio overviews.
What I use it for: I use NotebookLM in the research and discovery phase of client engagements. When a client sends over large volumes of internal documentation, policy papers, or financial reports, NotebookLM allows me to synthesise it rapidly before the first working session.
Example use case
For a public sector client, I loaded 14 internal process documents into NotebookLM and generated a structured summary of current-state operations in under 30 minutes. This replaced what would have been two to three days of document review.
Granola
What it is: An AI meeting transcription and note-taking tool that runs in the background during calls, captures everything said, and generates structured meeting notes automatically.
What I use it for: I use Granola for every client meeting. It removes the need to take notes during calls, produces accurate transcripts, and generates action items and summaries that can be shared immediately after the meeting. It integrates with Google Meet, Zoom, and Teams.
Example use case
After a 90-minute workshop scoping call, Granola produced a structured summary including agreed actions, open questions, and a timeline of discussion topics. This was shared with the client within five minutes of the call ending, replacing what used to be a 45-minute write-up.
Replit
What it is: A cloud-based development environment that allows you to build, run, and deploy software using AI-assisted coding with minimal setup.
What I use it for: I use Replit for rapid prototyping of client-facing tools, internal dashboards, and lightweight automations that need a user interface. It dramatically reduces the time from idea to working prototype, and is particularly useful for demonstrating what an automation could look like before committing to a full build.
Example use case
For a finance team, I built a working prototype of an invoice exception dashboard in Replit in under three hours. The prototype was enough to get stakeholder sign-off on the approach before the production build started.
V0 (Vercel)
What it is: An AI-powered UI generation tool that creates fully functional React components and web interfaces from a text description.
What I use it for: I use V0 when a client needs a polished front-end quickly. It produces clean, production-ready component code that integrates with Next.js projects. For client-facing portals, reporting dashboards, and internal tools, V0 removes the design bottleneck entirely.
Example use case
For a management reporting tool, I used V0 to generate the complete dashboard interface from a written brief in one session. The components went straight into the production codebase without significant modification.
Make.com
What it is: A visual automation platform for building multi-step workflows without code.
What I use it for: I use Make.com as the default orchestration layer for most client automation projects. Its visual canvas makes it easy to hand over to clients post-implementation, and it handles complex branching logic cleanly.
Example use case
For a legal services client, I built a Make.com scenario that triggers when a new matter is opened in their case management system. It automatically creates a document folder, generates a client welcome email from a template, and logs the matter in the CRM without anyone touching a keyboard.
n8n
What it is: An open-source, self-hosted automation platform with a code-capable workflow builder.
What I use it for: I use n8n when clients need self-hosted deployment for data sovereignty reasons, or when the automation requires custom code at certain steps. It is more technical than Make.com but more flexible.
Example use case
For a financial services firm, I deployed n8n on-premise to process incoming invoice data, validate it against their ERP, flag exceptions, and route approvals. The self-hosted requirement was non-negotiable due to GDPR obligations around financial data.
Microsoft Copilot Studio
What it is: Microsoft's low-code platform for building custom AI agents and copilots within the Microsoft 365 ecosystem.
What I use it for: I use Copilot Studio when clients are heavily invested in Microsoft 365 and want AI capabilities that live inside Teams, SharePoint, or their ERP. The native integration with Microsoft systems removes the need for third-party connectors.
Example use case
For an operations team, I built a Copilot Studio agent embedded in Teams that answers questions about internal processes, routes HR queries to the right person, and pulls live data from SharePoint lists. It handles around 40 percent of internal queries without human intervention.
UiPath
What it is: An enterprise RPA (robotic process automation) platform for automating desktop and web-based tasks at scale.
What I use it for: I use UiPath for automation involving legacy systems without APIs, desktop applications, or high-volume repetitive tasks that need enterprise reliability and audit trails.
Example use case
For a manufacturing client with a legacy ERP that offered no API access, I used UiPath to automate daily stock reconciliation: the bot opens the system, extracts the report, compares it to the warehouse management system, and flags discrepancies to the operations manager. No code changes to the ERP were needed.
Zapier
What it is: A widely-used automation platform focused on simplicity and breadth of app integrations.
What I use it for: I use Zapier for simple trigger-action automations where a client already has a Zapier account or needs a quick win without complex setup. For anything beyond three or four steps, Make.com or n8n is more appropriate.
Example use case
For a marketing agency, a basic Zapier automation routes new Typeform submissions into HubSpot, sends a Slack notification to the account team, and creates a folder in Google Drive. Simple, reliable, and set up in under an hour.
Industry Use Cases
I have real project experience across five sectors. Here is what the automation problems look like in each, and the tools I typically deploy.
Financial Services
Finance teams spend a disproportionate share of their time on data entry, reconciliation, and report generation. I typically automate invoice processing from email or shared drive intake through ERP posting, with exception flagging. Compliance reporting that used to take a day runs in minutes. Tools: n8n (for self-hosted data requirements), UiPath (for legacy ERP interaction), Claude (for document extraction from unstructured invoices and statements).
Healthcare
The highest-value automations in healthcare are around administrative workflows: referral processing, appointment scheduling, patient communications, and clinical document summarisation. AI handles the volume; clinicians handle the judgment. Tools: Claude (for reading and summarising clinical documents), Make.com (for appointment workflow orchestration), Copilot Studio (for internal team queries in NHS-adjacent environments).
Education
Education organisations deal with high administrative overhead across enrolment, parent communications, and curriculum management. I build pipelines that automate enrolment status updates, generate personalised parent communications based on student data, and support curriculum research with AI-assisted content synthesis. Tools: Make.com, ChatGPT/GPT-4, Zapier (for lightweight CRM and email integrations).
Legal Services
Legal AI automation is predominantly around document intake, matter routing, contract review pre-screening, and client communication. The key constraint is accuracy: errors in legal documents carry real consequences. AI is used to surface, flag, and route, not to make decisions. Tools: Claude (for document reading and summary), Make.com (for workflow orchestration), UiPath (for legacy case management systems without APIs).
Sales and Revenue Operations
Sales ops automation has the fastest visible ROI. Lead enrichment, follow-up sequencing, CRM hygiene, and pipeline reporting can all be automated with minimal disruption. The result is a sales team that spends more time in conversations and less time doing admin. Tools: Make.com, GPT-4 (for personalised outreach generation), HubSpot or Salesforce native AI features where available.
What Does an AI Automations Manager or Consultant Actually Do?
This is one of the most searched questions in the AI automation space, and it rarely gets a direct answer.
An AI automations manager or consultant does five things:
Maps your business processes
Identifies which tasks are manual, repetitive, and high-volume. Documents the current workflow before designing the automated version.
Selects the right tools
Matches the automation platform to the job. Not every problem needs the same solution. The tool selection phase saves significant implementation time.
Builds and tests the automation
Configures the workflow, connects the systems, handles exceptions, and tests edge cases before anything goes live.
Integrates with existing systems
Connects the automation to your ERP, CRM, email, document management, or communication tools without replacing them.
Trains and hands over
Ensures the team can run, monitor, and extend the automation independently. Dependency on the consultant is a failure mode, not a business model.
How to Get Started with AI Automation in Your Business
The most common mistake is starting with the tool rather than the problem. Here is the right sequence:
Identify your highest-cost manual tasks
Ask every team lead: what do you or your team do every week that is repetitive and time-consuming? Collect the answers. The overlap is your starting point.
Pick one process to automate first
Start with something high-volume, low-complexity, and self-contained. A process that involves one input, one transformation, and one output is ideal for a first automation.
Map the process before building anything
Document every step of the process as it currently happens. Every exception, every edge case, every system the data touches. Automation built on an incomplete map fails at the edges.
Choose the right tool for the job
Use the tool comparison above as a starting point. If in doubt, Make.com is the right default for most SME and mid-market projects.
Measure the result against the baseline
Record how long the process took before automation. After go-live, measure the same metric. The difference is your ROI. This is the number that unlocks budget for the next automation.
Frequently Asked Questions
Make.com vs n8n vs Zapier: which should I use?
Zapier is the simplest but expensive at scale and limited for complex workflows. Make.com handles multi-step, conditional automations well and is accessible to non-technical users. n8n is the most flexible, ideal for technical teams or where self-hosting is required. For most businesses: Make.com. For developer-led teams or data sovereignty requirements: n8n.
Is UiPath or Make.com better for finance automation?
It depends on your ERP. If your finance system has an API, Make.com is simpler and faster to deploy. If your ERP is legacy and offers no API access, UiPath's RPA approach works by automating the user interface directly, which requires no ERP changes. Most modern implementations combine both: UiPath for the legacy system interaction and Make.com or n8n for orchestrating the downstream steps.
What does an AI automation consultant charge?
Project-based engagements for a single workflow automation typically start from £5,000 to £15,000 depending on complexity and the number of systems involved. Ongoing retainer support for multiple automations runs from £2,000 to £5,000 per month. A free automation assessment will give you a realistic scope and cost estimate before any commitment.
Do I need a technical team to work with an AI automation consultant?
No. The consultant handles all technical configuration. What you need from your side is a good understanding of the current process (who does what, in what system, in what order) and an engaged stakeholder who can answer questions during the mapping phase. The handover training is designed for non-technical users.
Consulting enquiry
Book a free 60-minute automation assessment. We map your processes, identify your highest-value opportunities, and give you a realistic starting point.
Get Free AssessmentTraining enquiry
We offer hands-on AI training workshops for operations, finance, HR, and sales teams. Practical skills your team can use from day one.
Enquire About Training