Back to Insights

Your AI tools sit next to the work, not in it

TL;DR

  • Buying a tool does not move the work. The work stays where it was.

  • Most AI sits beside the work as one more tab to open and check.

  • Payoff shows up when the tool lives inside the step where work happens.

  • Start with one follow-up that keeps slipping. Put it inside one place.

  • Give agents small authority first. Grow it only after they earn trust.

You bought the tools. The follow-ups still live in your head.

That is the pattern I keep seeing. A new app gets added. A demo looks good. Two weeks later the lead is still chased on WhatsApp, the number is still fixed in Excel, and you are still the person who remembers who promised what to whom.

The tool did not fail. It never got inside the work.

Why do follow-ups still slip after buying new tools?

Follow-ups slip because the reminder lives outside the moment where the work happens. A lead arrives in one place. The chat happens somewhere else. The next step sits in your memory.

So the chain breaks in the gaps. Someone gets busy. The chat scrolls up. The promised callback was never written where the next person can see it. Then the prospect goes cold, not because your product was wrong, but because nobody circled back in time.

I see the same shape in my own work. If a task depends on me remembering to check a second tab, it will slip on a busy day. Not sometimes. Every busy day. The fix is not trying harder to remember. The fix is putting the next step where the work already sits. That is the whole idea behind BaseEngine: the system holds the follow-up so a person does not have to.

What does it mean for AI to sit beside the work?

Sitting beside the work means the AI is one more place you have to go. You finish a call, then you open a separate tool, paste in the notes, wait, copy the output back, and then do the actual follow-up yourself.

Nothing connects on its own. Every step needs you as the bridge. That is why it feels like extra work. Because it is extra work. You now do the old job plus the work of feeding the new tool.

Research on large companies shows the same split. PYMNTS Intelligence tracked 60 enterprise finance chiefs and reviewed 437 AI deployments (PYMNTS Intelligence, September 2026). Only about one in five deployments, 20%, were actually embedded inside a workflow. The rest sat next to the work as broad or early use.

  • Beside the work. You open it, feed it, copy from it. Used in week one, quiet by week four

  • Inside the work. It is there at the step you already do. Becomes the default way the step runs

This is not a small-company problem or a big-company problem. It is a placement problem. You can read the full findings in the PYMNTS report summary and in Karen Webster's take on the agentic enterprise.

Why does work default back to WhatsApp and Excel?

Work defaults back because WhatsApp and Excel ask for nothing. No login. No training. No new habit. The chat is already open. The sheet already has the columns your team made up over time.

A new tool asks for a new habit before it gives any value back. Your team will not pay that cost on a busy Monday. They will use the fastest path in front of them, and the fastest path is the one they used yesterday.

So the question is not which tool is best. The question is whether the new way is easier than the old way at the exact moment of work. If it is not, the old way wins. Every time. This is why we build around how a business already runs when we do AI automation work. We do not ask a team to abandon its habits first. We put the system where the habits already are.

What changes when a tool lives inside the workflow?

When a tool lives inside the workflow, the lead, the follow-up, the quote, and the payment note sit in one place the team already opens. Nobody asks you for an update because the update is visible. Nobody digs through chats to answer what happened with this customer.

The PYMNTS numbers point the same way. About 25% of firms said AI had paid off overall. That rose to 55% where one or two functions were embedded in daily work, and to 93% where three or more were embedded (PYMNTS Intelligence, September 2026). Read that carefully. The study shows a link, not proof that embedding caused the returns. But the step from 25% to 93% is too big to ignore.

Enterprises in that study used AI in an average of seven of eight business functions. Adoption was wide. Embedding was rare. That is the gap. Not access to tools. Tools are everywhere. A place inside the workflow is the rare thing.

Here is what inside looks like for a small firm:

  • A lead lands and already has an owner and a next step. No pasting.

  • The follow-up date lives on the lead, not in someone's chat.

  • The quote and the chat history sit together, so a new person has context.

  • You open one view in the morning and see what is stuck and with whom.

We describe this operating shape on the BaseWorks homepage and in more detail on the BaseEngine page. The point is simple. One place. One owner. One next step.

How much authority should you give an agent on day one?

Give it a small job with hard edges. Let it draft. Let it remind. Let it prepare. Keep approval with you for money, for anything you publish, and for anything you cannot undo.

That is how I run my own setup. The routine work moves on its own. Anything that spends money, posts in public, or cannot be taken back stops and asks me first. Trust grows because the edges hold, not because I hoped they would.

The enterprise research makes the same point in bigger words. Authority has to be earned through experience, with clear limits on amounts, people, and actions. A board, or a founder, should be able to explain what the agent was allowed to do and why. If you cannot explain it, the limit is too loose.

Start narrow. Watch ten runs. Fix what it gets wrong. Then widen one edge at a time. Boring. That is why it works. If you want the longer story of how we think about this, the Insights page is where we keep these notes, and the About page covers how BaseWorks works with founders.

Where does your work actually live today?

For most small firms, work lives in a chat thread, a sheet, and the founder's memory at the same time. Not where the tool says it lives. Where does it actually live at 6 pm when a customer asks for an update?

If the answer is a chat thread, a sheet someone named final_v3, and your memory, then your work lives in three places at once. That is why it slips. No tool fixes that. Moving the work into one place fixes that.

Pick one follow-up that slipped this month. One. Ask where it should have lived so the slip could not happen. That answer is your first build. Not a platform. Not a migration. One follow-up, one owner, one place.

That is the conversation I want this piece to start: where does your work actually live, and what is the one follow-up you are tired of holding in your head?

Want to think it through for your business? Start with the BaseEngine overview and see if the shape fits how you already work.

Frequently Asked Questions

Why do follow-ups keep slipping even after we bought a CRM?

Follow-ups slip because the next step lives outside the place where work happens. The lead sits in the CRM, the chat sits on WhatsApp, and the reminder sits in someone's memory. The gap between those three is where callbacks get lost, especially on busy days or when someone is on leave.

Why does my team go back to WhatsApp and Excel?

Your team goes back because those tools ask for nothing new. No login, no training, no changed habit. A new system asks for effort before it pays back. On a busy day, people pick the fastest path in front of them. The new way has to be easier at the moment of work, or the old way wins.

What is the difference between AI beside the work and inside it?

AI beside the work is a separate place you open, feed, and copy from. You stay the bridge between every step. AI inside the work sits at the step you already do. The lead already has an owner, the note is already attached, and the next step is already visible without extra effort from you.

Do I need to replace our tools to fix this?

No. Start with placement, not replacement. Pick one follow-up that keeps slipping and give it one home, one owner, and one visible next step. Most teams do not need more software. They need the work they already do to live in one place their team already opens every day.

How much should I let an agent do without asking me first?

Let it handle routine drafts, reminders, and preparation on day one. Keep your approval on money, public posts, and anything that cannot be undone. Watch a small set of runs, fix the errors you see, and widen its room one step at a time only after it earns trust.

What should I fix first if work lives in my head?

Fix the one follow-up you are most tired of remembering. Give that single task a home in one system, a named owner, and a date the whole team can see. Small and finished beats large and planned. Once one loop stops depending on your memory, the next loop gets easier to move.

By Aravind Bhati, Founder, BaseWorks