Where to start with AI

Every company we talk to has the same story. They bought the automation tools and then did not know what to do with them. They did not know where to start. That is completely normal.

Whether you are vibe-coding your automations with LLMs or running them on n8n, Make, Zapier, or even just Claude, you likely have a hard ceiling on how much impact you can make. The reason is simple. You do not know how your company actually runs day to day. You do not have the visibility you need into the real work. That is where you need to start to make any of this work.

We have seen this play out dozens of times. A team picks something flashy to automate, it saves twenty minutes a month, and everyone quietly loses interest. The fix is boring but it works. You have to actually look at the work before you try to improve it.

See the work first

Most companies have no idea how their teams actually spend their time. They have job descriptions and org charts and maybe a few SOPs that nobody has updated since 2022. None of that tells you which tasks eat three hours a day or which workflows involve seventeen tab switches between five different apps.

You need an accurate picture of the real work before you can make any good decisions about what to automate, what to simplify, and what to leave alone. There are three ways to get that picture, and they are very different in cost and reliability.

Option 1: Do it yourself

Sit down with your team. Interview people. Watch them work. Write the processes down in a document or draw them in something like Miro or Lucidchart.

This rarely works well. People cannot accurately describe work they do on autopilot. The invoice downloads, the status update emails, the copy-paste between the ERP and the spreadsheet. These tasks happen hundreds of times a week and nobody mentions them in an interview because they barely notice them anymore.

Manual process mapping is also brutally time intensive. A single department can take weeks. And the initiative almost always loses momentum before it finishes. The person running it gets pulled into other priorities, the flowcharts go stale, and the whole effort quietly dies. We have watched this happen at companies of every size.

If your team is five people with straightforward workflows, it can work. For anything bigger, the effort almost never pays off.

Option 2: Consultants

Hire outside help. There are plenty of firms that specialize in process mapping and operational efficiency. At the enterprise level, the Big Four (Deloitte, EY, PwC, KPMG) all offer this, often under their "intelligent automation" or "digital transformation" practice.

These people know how to run process discovery projects. They have frameworks, they have done it many times, and they deliver a polished report with clear findings.

But it is expensive. A meaningful engagement starts at tens of thousands of dollars and goes up fast. Gartner estimates that the average process mining initiative costs $250K+ at the enterprise level. And even with experienced consultants, the method is still mostly interviews and observation. More structured than doing it yourself, but it still depends on people remembering work they do without thinking about it.

If you have the budget and leadership backing, this path delivers. For small and mid-size teams, it is simply out of reach.

Option 3: Automated discovery

Five years ago this option did not exist. AI has gotten good enough to watch how work happens on a screen, recognize the patterns, and map them without anyone having to describe anything.

We built MemoryLane to do exactly this. It runs in the background on your team's machines, watches the real pattern of apps, windows, and repeated steps over a normal work week, and produces a ranked list of the workflows worth automating first. No interviews. No flowcharts. No six-figure consulting bill.

You are not relying on what people remember. You are looking at what actually happened. Every tab switch, every repeated sequence, every chunk of time spent in a specific app. The data is there whether someone thinks to mention it in an interview or not.

Setup takes about five minutes per machine. Your data stays on the device. Nothing leaves unless you choose to share it, and even then, everything runs through zero-data-retention endpoints. Within a week you have a picture of where time goes that would take months to build any other way.

The enterprise process mining tools like UiPath, Celonis, and Mimica do similar things, but they are priced for large organizations and require significant integration work. MemoryLane starts at $50 a month and works out of the box.

Getting your team onboard

If you want to push AI adoption in your company, you need people onboard. The best way to do that is to lead by example.

Start with yourself. Run the tool on your own machine, look at the results, and share what you found with the team. When an executive or team lead shows their own data and says "look, I spend four hours a week on this and we can automate it," it removes the fear. People see a concrete example of the value instead of a vague pitch about AI.

Then find two or three champions. People who are already excited about automation and want to try it. Give them access and let them run it. When they showcase how they benefited, it brings others along naturally. We see significantly better adoption when it spreads this way compared to a top-down mandate that nobody asked for.

Frame the rollout as the company letting AI handle the boring manual work so people can focus on work worth their time. That message matters. Nobody wants to hear "we are monitoring your workflows." They want to hear "we are finding the stuff that wastes your day so we can get rid of it."

Start small

Once you have the ranked list of workflows, pick one. Just one. Whichever tool gave you the visibility already tells you how long it takes and how often it happens. That is your baseline.

Build the smallest automation that actually helps. One step taken off someone's plate, done reliably. Run it for a week and compare against that baseline.

If it holds up, you have something better than a time saving. You have proof that the approach works. And proof is what turns a single win into a budget, a roadmap, and a team that actually wants to do the next one.

Next steps

Start by getting visibility into how work actually happens. Try MemoryLane on the self-serve plan and get in touch if you need help with a more customized rollout, security and data controls, analysis, or automation.