The Quiet Death of the AI Tools That Worked

By Nasly Duarte

Last night I was at CEDIA, the smart home trade conference. The conversation I want to write about happened on the exhibit floor. It ended with a tool I love announcing its own shutdown in my hand.

The Conversation

This booth caught my eye because i read homes and assumd they were building smart home from scratch.. 

I was talking to Philippe Lafoucrière, Founder and CTO of Selora Homes, and Z. Michael Miranda, also at Selora. Selora is an AI-first smart home company built on Home Assistant. Philippe sold his previous company to GitLab in 2018. He is not a man who needs a pitch.

They were demoing their software to me. The whole demo was on a phone. Philippe and Michael kept tapping through screens. Features. Systems. Tedious settings.

Micheal even suggested Philipp show me the automated suggestions the app gives you. 

I watched them looking down at the phone over and over.

The Realization

And I realized something while it was happening.

Every morning when I get ready, I put Huxe on. I do not touch my phone. The audio just plays. I move through my morning. I got down notes and blog post ideas. 

Smart home software is supposed to free people from their devices. It is being demoed on the device. The whole industry is making people look at phones to control the things that are supposed to make them not need phones.

I told Philippe and Michael. Voice and audio, not screens. Their tech is the home. Huxe is the way you experience it without holding a phone. The integration would be real.

They asked How? 

I explaind there is this app called Huxe. It's a brief app and i went to my phone to show them (ironically) and went to the app. 

What Happened Next

I wanted to show them Huxe. The exhibit floor was loud. I opened the app, started a playback, and handed the phone to Michael so he could hear it.

He held it to his ear. He looked confused.

He told me he just heard the app was closing.

I took the phone back. I replayed it. I confirmed it for myself.

I was mortified.

The pitch I was making in that moment died on the table while I was making it. Three of us at CEDIA. The CTO of an AI-first smart home company, his commercial counterpart, and me. The shutdown announcement played in real time. The tool I was pointing to as the missing piece announced itself out of existence in the same breath.

The Pattern

Huxe was an AI-powered audio briefing app built by the team that originally built NotebookLM at Google. You told it what you cared about. It generated audio about it. You listened while you got ready in the morning. The phone stayed in your pocket.

Spotify shipped a similar personal-podcast feature the day before Huxe announced the shutdown. Adobe, Amazon, ElevenLabs, and Meta have all emulated the underlying capability. The consumer AI market commoditized the feature in under a year.

The official statement from the Huxe team was short. They chose to wind down. They are moving on to new things.

But they had options. A paid tier. A customizable version. A B2B pivot. Users like me would have paid serious money for the calibrated audio briefing.

They chose not to take them.

The Disconnect Argument

The conversation I was trying to start with Philippe and Michael was about voice-first interaction with a smart home. The reason the demo on the phone bothered me was that it modeled the wrong relationship between human and house. If your home is supposed to be smart, you should not need to look at a screen to talk to it. The interface should disappear.

Phones build disconnect. Not because phones are bad. Because every interaction with a phone is a moment where you are not in the room you are standing in. You are not present with the person next to you. You are not noticing the morning.

The tools worth fighting for are the ones that get out of your way. The ones that give you back the morning instead of taking it. The ones that make the phone unnecessary instead of central.

The Sharp Version

We lost a tool like that this week. Not because users did not want it. Because the team chose to stop instead of building the paid customizable version users would have paid for.

This is the pattern. Small builder makes something good. Big platform copies the feature. Small builder folds. The user — the person who actually integrated the product into a working day — has no vote in any of it. Not because we lack interest. Because no one builds the path.

The next builder who makes something I love this much, I am paying for. Even if it is free. Especially if it is free. Because what I want to keep is the option to keep it.

For Members

I wrote a longer version of this post for Mindful Dollar members on Buy Me a Coffee. It includes the part of the story I left out here — what Huxe actually did inside my mornings, why losing it is a real operational setback for the work I am building, and what I think the next builder needs to get right.

If this post resonated, the deeper version is where the real reflection lives.

Read the members post on Buy Me a Coffee →

What Gets Lost Between Design and the General Contractor

Last night I was at CEDIA, the trade conference for the custom integration industry. The conversation that stayed with me was with Joseph Kolchinsky, Founder and CEO of OneVision Resources. Joseph has spent more than twenty years building a service platform for custom integrators in the smart home industry.

In this post I share what he taught me about the communication deficit between design and the general contractor, why entire industries get built around problems that no one closes, and where the research has to start.

 

The Question I Asked Him

I asked Joseph one direct question. As an AI student and an operations accountant, what value can I bring to this industry?

He answered with two words.

Processes and workflows.

 

The First Problem I Raised

I raised the silent modification problem first. The line items that change between the estimate and accounts payable. The work that gets rebilled under a different number than the one approved.

Joseph did not see the point of it. His response was direct.

 Its normal for material and labor to change. 

He was not dismissive. He was honest. He has watched that pattern play out across thousands of integrator projects and he has accepted it as a structural feature of the industry. I read the room and moved to a different question.

I asked him about the communication deficit between design and the general contractor.

He agreed it exists.

CEDIA brings together the custom integration industry that absorbs what design and construction leave undone.

What the Deficit Actually Is

An architect produces a design. A general contractor builds it. Between those two events, information has to travel. Drawings. Specifications. RFIs. Addenda. Change orders. The intent behind each decision. The reasoning behind each constraint.

Some of that information moves intact. Some of it gets dropped. Some of it gets withheld. Some of it gets changed.

By the time the building reaches the people who actually have to construct it, the design intent is partial and the reasoning behind the constraints is mostly gone.

Who Absorbs the Gap

The general contractor absorbs the gap. They build through it. They invent the workarounds. They eat the cost of every upstream fidelity loss in their schedule, their margin, and their reputation.

The custom integrator sits one step further downstream. They install the low-voltage systems, the audio-video infrastructure, the smart home technology. They show up to a building that was built from a design that already lost information twice. The wiring path they need was not planned. The conduit they need was not run. The infrastructure they need was not specified.

They invent another workaround.

This is the part Joseph has spent twenty years inside of.

What OneVision Actually Does

OneVision is a service platform that supports custom integrators after the work is sold and the project is underway. The platform exists because the deficit exists.

Integrators need ongoing technical support precisely because the upstream chain hands them incomplete information. The service layer absorbs what the design and construction process did not transmit.

Joseph has built a real business at that seam. The seam should not exist. The industry has accepted that it does, and Joseph has built a company that makes it more livable for the people working downstream of it.

This is what I call the workaround economy. An entire service layer that exists because of a structural problem upstream, profits from it, and over time becomes a constituency with its own reasons to keep the problem open.

The honest read on Joseph's business is not a criticism. It is a description. A successful operator looks at a structural inefficiency and builds something that absorbs the cost of it. OneVision absorbs the cost. It does not close the gap. Neither do the BIM coordinators, the design-build firms, the construction managers at risk, the claims consultants. Every node in the workaround economy is doing the same thing in a different shape.

What He Told Me About Closing the Gap

After we talked about the deficit, I told him I wanted to change it.

He pushed back. The push was not against the idea. The push was against the approach.

Walking into an entrenched industry and demanding changes is not the entry point. Architects will do what architects have always done. General contractors will work around what they have always worked around. Outsiders who arrive with fixes get filtered out by the same workarounds that keep the system running.

He was right.

Change in a legacy industry does not start with a fix. It starts with making the existing system legible. The workarounds that look broken from the outside are often doing real work. The patterns that look like dysfunction are often informal contracts that hold the project together when the formal contracts cannot.

An outsider has to understand what each workaround is doing before they can propose anything that would survive contact with the industry.

The custom integration industry sits one step downstream of every design-to-GC handoff.

Where the Research Has to Start

The communication deficit between design and the general contractor is not a documentation problem. It is not a technology problem.

It is a structural problem about how information loses fidelity when it crosses an organizational boundary that was not designed to preserve it.

The research that matters would measure the deficit. Not in surveys. In the actual artifacts.

The Five Artifacts Where Fidelity Drifts

• The design intent at the architect's desk. The reasoning, constraints, and decisions that shape the initial drawings.
• The construction documents. What gets formalized for permitting and bidding.
• The RFI log. Every question the GC had to ask because the documents did not answer it.
• The change order. Every modification made during construction, with whatever rationale was captured.
• The as-built drawings. What was actually built, handed to the integrator who installs everything else.

The drift between those five artifacts is the deficit. No one has measured that drift cleanly.

The industry has built workarounds for the consequences and a service economy on top of the workarounds. Joseph's company is one node in that service economy. There are hundreds of others.

The research question is whether the drift can be measured at all, and whether the act of measuring it changes the behavior at any point in the chain.

That is the question I would want to spend a few years answering.

The Bottom Line

Joseph told me two things in the same conversation.

The first was that estimate-to-AP modifications are how it has always been. The second was that processes and workflows are the value an AI student and accountant can bring to the industry.

Both statements are correct. They are also in tension. You cannot improve the processes and workflows without addressing the modifications they currently absorb.

That tension is the work.

The communication deficit between design and the general contractor is one place where it lives. The estimate-to-AP gap is another. Every legacy industry has its own version. Architects and general contractors are not the villains. Custom integrators are not the heroes. The workaround economy is the system. The system is doing what it was built to do.

The question is whether anyone wants to design something different.

I am not proposing one today. I am pointing at the seam.

Accounting Saw

For a deeper dive on this topic. Read my work on this and where I see things need to change by joining my BMAC Community. 

 

Link: buymeacoffee.com/girlgoneverde/what-gets-lost-between-design-general-contractor

---

Nasly Duarte is the founder of Mindful Dollar and a student in the Bachelor of Science in Applied Artificial Intelligence program at Miami Dade College. She writes about operations accounting, autonomous business systems, and the financial architecture that holds them together. 

If you find value in these breakdowns and want to support the work I do, bringing you the latest in AI, operations, and business systems, please consider treating me to a virtual caffeine boost! You can hit the "Buy Me a Coffee" button below, and yes, I am officially accepting crypto now too! Your support keeps this whole thing running.

 

Space Intelligence App

Space Intelligence 

You haven't heard from me in a few weeks. Here's why.

I've been building something called SoulAccess.

It's a phone-based platform that gives members secure access to sacred space outside of service hours. Faith centers maintain complete control. Members access the buildings their tithes already sustain, at the hours they actually need them.

The short version: most churches, mosques, and temples sit empty more than 90 percent of the week. Meanwhile people experience grief, anxiety, panic, and crisis at the hours when those buildings are locked. SoulAccess is the access layer that closes that gap.

I started building it on the National Day of Prayer  weeks ago. Since then:

The landing page is live at soulaccess.netlify.app. A high-fidelity clickable prototype lives on the same page. Scroll down to "Try it yourself." The prototype is fully bilingual. English and Spanish. Pilot conversations are underway with three pastors and the Mayor of Hialeah. A spot at the next GDG South Florida workshop is locked in to demo it with Misfit Labs. The whole thing is open source under CC BY 4.0 at github.com/nasly-ai/soulaccess.

You are the community that made it possible for me to spend this kind of time building. Before I share this anywhere else, I want to share it with you.

Three things you can do if you want to be part of this.

• One. Visit the site. Click through the prototype. Tell me what lands and what doesn't. Honest feedback from this group is worth more to me than the polished feedback I'll get anywhere else.

• Two. Forward this to someone. A pastor who has been on your mind. A friend who runs a community space. A developer who cares about this kind of work. The hardest part of building right now is reaching the right people. You can help just by sharing.
• Three. If you can, support this work financially. I don't draw a salary from any of this yet. Every coffee buys me another evening of building.

This is the work I'm most proud of doing. It's also the work that scares me the most. Both of those things being true at the same time tells me I'm in the right room.

Read the Full Post of my realization it was spatial not sapce intelligence. lol your live and you learn.  https://buymeacoffee.com/girlgoneverde/spacial-intelliegence 

Is AI Coming for Accountants - National Accountant Day

I truly believe AI Is Coming for Accountants. Here Is the Distinction That Will Matter More Than Any Credential.

The accountants who are afraid of AI are asking the wrong question.

The question is not whether AI will replace accountants. The question is which accountants AI will replace first, why the answer has nothing to do with credentials, years of experience, or the software you know how to use.

I have spent 20 years inside accounting operations across construction, manufacturing, insurance restoration, retail, and trade services. I have watched skilled accountants get bypassed, automated around, and made redundant — not by AI, but by anyone who understood the system underneath the numbers better than they did. AI is about to accelerate that pattern at a scale no one is fully prepared for.

Here is the distinction that will determine who survives the next three years of accounting automation  and it is not the one most people are talking about.

The Two Layers of Accounting

Every accounting function has two layers. Most accountants only know one of them.

Layer 1 is the software layer. This is where most accountants live. QuickBooks, Xero, Bill.com, SAP, NetSuite. How to enter a transaction. How to reconcile an account. How to run a report. How to close the month. These are skills. They are real and necessary. But they are also exactly what AI is best at: pattern recognition, data entry, classification, reconciliation, and report generation at a speed and accuracy no human can match.

Layer 2 is the systems layer. This is where the accountant understands why the numbers look the way they do. How the chart of accounts was designed and whether it actually reflects how the business makes money. Where the transaction originates before it ever reaches the accounting software. Why the AP balance is wrong even though every invoice was entered correctly. What the cash flow statement is actually telling you about the next 90 days of the business.

Layer 1 is what AI replaces. Layer 2 is what AI cannot  "Yet".

The accountants who only know Layer 1 are already at risk. The ones who understand Layer 2 are not just safe. They are about to become significantly more valuable.

What Layer 2 Actually Looks Like in Practice

A Layer 1 accountant looks at an accounts payable aging report and sees invoices. A Layer 2 accountant looks at the same report and sees a procurement workflow that is three approvals deep, a vendor relationship that has been deteriorating for six months, and a cash position that will be short in 47 days if three specific invoices clear at the same time.

A Layer 1 accountant reconciles the bank statement. A Layer 2 accountant notices that the reconciling items are always in the same cost code, asks why, and finds a data entry pattern that has been masking a billing error for eight months.

A Layer 1 accountant closes the month. A Layer 2 accountant reads the month and tells the business owner what happened, what it means, and what decision they need to make before the next period opens.

AI can do Layer 1 faster than any human. AI cannot yet do Layer 2 — because Layer 2 requires understanding the operational context that produced the numbers, not just the numbers themselves.

Why This Distinction Matters More Than Your CPA

I say this with respect for the credential. The CPA is a rigorous, meaningful qualification. It tests technical knowledge at a depth that matters. But the CPA tests Layer 1 proficiency — knowledge of the rules, standards, and procedures that govern financial reporting.

It does not test whether you understand the business system that produced the financial data you are reporting on.

The accountants who will thrive in an AI-augmented world are the ones who can walk into any business, read the operational reality through the financial statements, and design the system that makes the numbers accurate, timely, and decision-ready. That skill is not tested on any exam. It is developed over years of being embedded inside operations — not just recording them.

AI is going to make the credential table stakes. Everyone will have access to technically accurate financial reporting at near-zero cost. The competitive advantage will belong to the accountant who understands the system well enough to know when the technically accurate report is operationally misleading.

What AI Actually Does to the Accounting Profession

AI does not eliminate accounting. It eliminates the parts of accounting that never required human judgment in the first place.

Invoice processing does not require human judgment. It requires pattern matching, data validation, and rule application. AI does that better than humans already.

Three-way matching does not require human judgment. It requires comparing three documents against each other at the line-item level. AI does that faster and more accurately than any AP clerk.

Bank reconciliation does not require human judgment in most cases. It requires matching transactions. AI does that in seconds.

What AI cannot do. What it will not do for a very long time, is understand why the three-way match keeps failing for one specific vendor, trace that failure back to a purchase order workflow that was never designed correctly, and redesign the system so it does not happen again.

That is a human judgment call. It requires operational knowledge, systems thinking, and the ability to see a financial problem as an operational problem in disguise.

The Accountant AI Cannot Replace

The accountant AI cannot replace is the one who sees the full system.

They understand that accounts payable is not just paying what is approved. It is the final step in a workflow that began at estimate or purchase order, passed through approval, touched procurement, and only reached AP after every upstream decision was already made. If AP is wrong, the problem is almost never in AP. It is somewhere upstream — and finding it requires tracing the transaction back through the system to its origin.

They understand that cash flow is not a report. It is a forecast built on the operational rhythm of the business — billing cycles, collection patterns, payment obligations, and the timing gaps between all three.

They understand that the chart of accounts is not a list of categories. It is an architectural decision that either makes the business's financial story legible or obscures it in ways that compound for years.

AI will augment every one of these capabilities. It will surface patterns faster, flag anomalies earlier, and generate reports that used to take days in seconds. The accountant who understands the systems layer will use AI to multiply their impact. The accountant who only knows the software layer will watch AI do their job and wonder what happened.

The Bottom Line

AI is not coming for accounting. It is coming for the parts of accounting that never required systems thinking in the first place.

The distinction that will matter more than any credential in the next three years is simple. 

Do you understand the system that produces the numbers, or do you only know how to record them?

That question has always mattered. AI is just about to make the answer impossible to hide.

#OperationsAccounting #AIforAccounting #FinancialSystems #AccountingAI #NaslyDuarte

Howdy! Nasly here. If you find value in these breakdowns and want to support the work I do bringing you the latest in AI, operations, and business systems, please consider treating me to a virtual caffeine boost! You can hit the "Buy Me a Coffee" button below, and yes, I am officially accepting crypto now too! Your support keeps this whole thing running.

Future PropTech Miami - Energy, Water and Space Intelligence

Energy, Water and Space Intelligence

By Nasly Duarte

Last week I attended an amazing insightful event. and i want to tell you about my trip to the Future PropTech Miami because my mind is completely blown! As an AI student I am always looking for the next big thing and this conference absolutely delivered. We need to talk about three incredible innovators that are completely revolutionizing how we use energy, water, and space.

First up is Akila. They are tackling energy waste using some seriously advanced AI technology. Specifically they rely on computer vision and machine learning algorithms. Imagine a digital twin of a building which is basically a perfect 3D virtual copy. Akila uses computer vision through a network of cameras and sensors to see exactly how energy is flowing and where it is being wasted in real time. Their AI software runs complex simulations to optimize cooling systems and lighting automatically. It is literally building an artificial brain for real estate!

Next I listened to Frederico Teixeira Egli talk about space intelligence. Have you ever wondered if a building is actually using its square footage properly? Frederico discussed how AI spatial analytics can find the absolute best use for a building footprint. By feeding foot traffic data and occupancy metrics into predictive algorithms the software can redesign floor plans to maximize utility. This means fewer wasted empty rooms and a much better flow of people. The AI learns how humans naturally move and adapts the space to fit our exact needs.

Finally we have Seth Guttenberg who is the CEO and cofounder of arkIQ. If you know me, you know that i can live with out electricity but I cant live without water. arkIQ built an incredibly smart water detection machine. They use internet-connected sensors combined with AI anomaly detection. The software learns the normal baseline water flow of a building. The exact moment a pipe starts leaking or water is being wasted the AI flags the anomaly before a catastrophic flood happens. It is predictive maintenance at its absolute finest saving millions of gallons of water!

We really need to pay attention to these technologies right now. Whether you are a solopreneur, a creator, or managing a massive operation, if we do not start looking at how we use energy, water, and space in both our commercial buildings and our own homes we are heading toward a very unfortunate future. The tools are here. AI is giving us exactly what we need to build smarter systems. 

Let us start building them today!

References

https://futureproptechmiami.com/conference/speakers

https://www.akila3d.com/blog/akila-profiled-in-nvidias-physical-ai-for-smart-cities-video/

Hey everyone! Nasly here. If you find value in these breakdowns and want to support the work I do bringing you the latest in AI, operations, and business systems, please consider treating me to a virtual caffeine boost! You can hit the "Buy Me a Coffee" button below, and yes, I am officially accepting crypto now too! Your support keeps this whole thing running.

AI, Process Replication, and Financial Operations

I DO NOT want AI to find accounting errors faster.

I want AI to help design systems where fewer errors make it to the financial statement

By Nasly Duarte

That is the difference between using AI as a cleanup tool and using AI as a financial operations architecture tool.

A lot of the conversation around AI in accounting focuses on speed. Faster reconciliations. Faster variance analysis. Faster anomaly detection. Faster close cycles.

All of that matters.

But if the process is weak, speed alone can become dangerous. A faster tool on top of a broken workflow does not create better financial visibility. It just moves the confusion faster.

The question I keep asking is this:

Will AI help business owners and operators replicate their real processes, or will it distract them with more fancy tools?

Because the financial statement should not be a mystery we solve at month-end.

Revenue should be traceable from customer activity. Costs should be traceable from labor, materials, vendors, usage, fulfillment, and delivery. Expenses should be traceable from commitments, approvals, invoices, payments, and allocations. Cash movement should connect back to the events that created it.

When that structure exists, reconciliation becomes a validation layer instead of the first place we go looking for the truth.

That is the work I am interested in as an AI Financial Operations Architect.

Not AI for the sake of AI.

Not another dashboard that looks impressive but does not explain the business.

Not automation that hides weak assumptions under a cleaner interface.

I want to help build systems where the business can explain how work becomes numbers.

That starts with line items, estimates, budgets, source data, controls, and process discipline.

If the business cannot explain how the numbers should be created, AI cannot responsibly automate them.

But when the process is clear, AI can help map the workflow, test assumptions, flag missing inputs, identify exceptions earlier, and strengthen the path from operations to financial reporting.

That is the real opportunity.

I do not want AI to make reconciliation the hero. I want AI to help design financial operations where reconciliation confirms the truth, instead of discovering the problem too late.