Limestone Digital

Enterprise AI transformation at startup speed.

🟪

[Limestone Digital @LimestoneHQ]

Some patterns we're seeing across our engineering clients that tell you where AI adoption is actually heading:

1. Companies are done spending blind on AI. A few months ago, most were approving unlimited API budgets for OpenAI and Anthropic with no tracking on where the tokens went. Today they're asking us to build governance layers and monitor every dollar. One client had three teams running up five-figure token bills before a single production agent shipped. Now they want dashboards that show exactly what's AI-generated, what it costs, and whether it's actually faster than adding headcount. That means more diagnostic work for us and meaningful savings for them.

2. Companies are accepting the real timeline. A few months ago, leadership expected to go AI-native over a few workshops and a handful of tool subscriptions. Now they're calling us because the codebase is seven years old and can't absorb automation without a full foundation phase first. Their developers are using AI at 60-70% adoption, but nobody's tracking what's generated vs. human-written. QA hasn't been touched. There's no governance. Going AI-native means rethinking org charts, handoffs, and process ownership across every department. Not a software purchase.

3. Companies are walking away from the big firms. The Accentures and McKinseys are all pitching AI transformations. Our clients keep showing up saying the same thing: "We paid seven figures for a strategy deck and nothing shipped." This is why firms like ours have a market. We operate at the engineering layer while having the business sense to map a process end to end, then build the agents to automate it. A year ago we had to explain this on every first call. Now prospects open the conversation with it.

In the last 90 days we've taken calls with PE firms, law firms, healthcare companies, VPN providers, construction platforms, and retail tech businesses. Some are direct competitors with each other. Nearly all came through referral.

AI SaaS won't transform your company. Giving every developer a Copilot subscription and calling it a strategy won't either. The only path to real operational ROI is combining a team that learns your business processes on the ground with engineering capacity to automate every manual workflow worth automating.

We built Limestone Digital around this thesis ten years ago. AI made it urgent. If you're looking to transform engineering delivery or automate business processes with production-grade agents, visit https://t.co/EVAbzMDhnA.

[Mark Ajzenstadt @mardehaym]

Most enterprises are about to sign a dozen AI contracts they'll rip out within two years.

AI intake assistant.
AI prior-auth checker.
AI claims coder.
AI call summarizer.
AI renewal predictor.
AI support reply bot.

Each one demos well. Each one is also a separate contract, a separate login and one more vendor holding patient or customer data.

Most get squeezed out, for three reasons:

1. ChatGPT, Claude and Gemini already connect to email, docs, CRM and tickets, and they get better every quarter.

2. Teams with engineers build their own agents inside their own stack, because they want control of their data and permissions.

3. The systems where the work happens (the EHR, the billing platform, core banking, claims) are building AI into the workflow itself.

Number 3 decides it. An enterprise pays for the outcome: the claim that clears, the ticket that closes, the audit that passes. Whoever owns that workflow owns the outcome.

So before your enterprise signs the next AI contract, ask one question: does it run inside the system your team already works in, or next to it?

Next to it is the one you'll replace.

[Limestone Digital @LimestoneHQ]

Got an agent in production with gaps on this list? Book a discovery call with us:

https://t.co/40XwXrZ5vs

[Limestone Digital @LimestoneHQ]

12 ways AI agents fail in production, and the control that stops each one:

1. It runs forever.
→ Turn limits, timeouts and a spend cap per run, enforced in code.

2. It says it's done when it isn't.
→ Code or a person checks the result, and its own report counts for nothing.

3. It retries a write and charges someone twice.
→ An idempotency key on every tool that changes a record.

4. It reads a table that looks right and was retired last year.
→ An allowlist of authoritative sources and a freshness check on each one.

5. It gives a different number each time you ask.
→ Numbers that must never vary come from deterministic code.

6. It follows instructions hidden in a document.
→ Treat everything it reads as data, and keep privileged tools away from runs that read untrusted text.

7. It drowns in its own context.
→ Tools that filter and return only what the step needs.

8. It forgets what it was doing halfway through a long run.
→ Keep state in a file or database, and checkpoint so it can resume.

9. It picks the wrong tool.
→ Fewer tools, one job each, clear descriptions and evals on tool choice.

10. It gets worse after a model or data change and nobody notices.
→ Pin the model version, run regression evals on a schedule, and alert on override rate and on changes in the shape of the data it reads.

11. Its approvals turn into rubber stamps.
→ Ask a person only for high-impact actions, show the evidence with the request and sample the ones that got approved.

12. It does something you can't undo.
→ A named approver before any irreversible action, and a compensation path for everything else it can write.

Bookmark this and mark which ones your agent has no control for. Fix those before you add another feature.

[Limestone Digital @LimestoneHQ]

Want to put AI to work in your operations next?

Get our free guide: how to pick your first workflow, measure the results and scale what works, with 3 case studies from healthcare, insurance and commercial lending.

https://t.co/9NAJfWRb3t

[Mark Ajzenstadt @mardehaym]

An AI consultancy is somewhere right now showing a PE portco a 3-year Gantt chart.

Love the confidence. But it’s completely detached from reality.

In 18 months, AI went from niche pilots to running core back-office operations. When models evolve every quarter, a tool chosen in January is baseline standard by June.

Planning three years out in this landscape is a trap. The roadmap is obsolete before the project kickoff meeting.

By year two, you're funding answers to obsolete questions.

The market leaders operate differently. They pick a single workflow, ship in weeks, measure real business impact, and iterate.

At exit, PE buyers don't care about your multi-year vision, they care about current results. "We’re in Phase 1 of our three-year initiative" is a failing grade.

The playbook right now:
> 90-day sprints tied directly to existing KPIs
> A lean, empowered team with full authority to kill failing bets
> Embedded AI integrated into daily ops, not an over-hyped "program" with its own branding

You need tech leaders built for speed and imperfect information. Bureaucracy will choke this execution every time.

Are any of you still greenlighting 3-year AI roadmaps?

[Mark Ajzenstadt @mardehaym]

Acquisitions put vendor relationships up for review. New owners look at every contract, every partner and every line item.

When PTI Security was acquired by ASSA ABLOY, we stayed.

PTI has been building security and access control systems for self-storage since 1979.

Its StorLogix Cloud platform runs across 40,000 facilities. When they hired us in September 2021, they needed a platform rebuild.

We started with a product assessment. Before touching the code, we mapped out who used the platform, where it was falling short and what was missing from their workflows. Then we rebuilt StorLogix Cloud 2.0 from the ground up.

After the rebuild, they trusted us with the next problem, then the next. We built a tenant app to replace the legacy mobile experience, an installer tool for smart lock commissioning and an admin portal for managing the business side.

That first engagement grew into four products, built by a team of eight engineers. This was before we were using AI to write code, so there were no coding agents, copilots or generated code.

Every feature was built by a developer who understood the system and reviewed by another who challenged the work.

Their CTO, Nathan Davenport, gave us a 5.0 on Clutch. He called us at @LimestoneHQ “experienced, highly skilled, and adaptive” and said we “monumentally improved the platform's loading speed.”

Then came the acquisition. ASSA ABLOY reviewed what PTI had and kept our team on.

Five years later, the same team is still working with them, with four products built entirely by hand behind us.

Every buyer I talk to wants a partner they won't have to replace. Whatever tools you use, you still have to earn the next engagement.

[Mark Ajzenstadt @mardehaym]

If I was a PE fund about to pay a premium for an AI company, I'd open the codebase before I opened the model.

A PE firm asked us to look at an acquisition before close. The company had $4.2M ARR, growing 40% year over year, and wanted 12x revenue.

Management had 14 slides explaining its "foundational model architecture."

We opened the codebase. GPT-4o, a system prompt, a React frontend and roughly 600 lines connecting them.

The proprietary model described in the deck wasn't there. The buyer walked.

A company built on OpenAI can still be a valuable business. You just need to know which of those you're paying for.

Four questions before close:

01 → Model dependency If the model vendor changes pricing or retires the model, what happens to quality and margins? Have the team run an alternative against the same customer tasks.

02 → Training data liability Where did the "proprietary data" come from, and what do the licenses actually allow? Permission to retrieve is not permission to train.

03 → Talent concentration Ask someone other than the original builder to deploy it, investigate a failed run and explain how they test a change. If every answer requires the same person, that's a close issue.

04 → Integration failure rates Pick one workflow it would have to run inside your portco. Have engineers inspect the systems, estimate the work, and name the owner. Do that before counting the savings.

Get engineers with hands on the keyboard into the engagement during exclusivity.

[Limestone Digital @LimestoneHQ]

👀

[Mark Ajzenstadt @mardehaym]

The difference between an AI consultancy and a hands-on implementation firm is simple.

One sells a long transformation. The other ships work, gets first numbers on the board, and iterates from real metrics. That is the better path: implement, measure, adjust, move.

Not a year-long generic AI program with workshops, frameworks, and no operating data.

Every engagement we run at @LimestoneHQ has fewer people on it now than when it started. The firm is growing faster than it ever has.

A consultancy has to keep an expensive person busy.

$350 an hour, billed by the day. The work is built to last. Scope expands. The team expands. Phase two gets sold in month three. You end up with a $500K deck about “unlocking potential.”

Ask people two or three levels below the executive who signed. They can’t stand consultants. They watched the circus.

What the firm is really selling is cover: I hired a top firm, so I made the responsible choice. If it fails, I still hired a top firm and there's nothing else we can do.

For a mid-market company, that’s an expensive way to avoid deciding.

Consultants do earn it sometimes. A dynamic consultant at $5K an hour who unblocks the thing that’s been strangling operations for two years is worth every dollar. A specific, painful problem for the business solved. That’s a fair trade.

Generic advice is the trap. In 2026, if the advice isn’t tied to a real problem, you can open a model, ask for the top five strategy frameworks, give the output to your HR lead with deliverables attached, and start Monday.

You’ll learn more from the first failed iteration than from three months of workshops. If you need a $350-an-hour outsider to tell you what your strategy is, the honest move is to change your job.

Vendors are selling the same billable-day model with an AI sticker on it. When a mid-sized or large engineering shop says they’ve gone AI-native, I ask one question: how many people have you let go, and why are you still hiring more? If the answer is none and lots, they aren’t passing any efficiency to you.

They added “AI” to the rate card and kept the man-day model underneath.

The small firms are the newest version. A group assembles a team to ride the AI demand wave, hires people with case studies, and suddenly the firm has case studies. But there’s no history behind the logos.

Our @LimestoneHQ AI velocity pod is one senior engineer, agents across the whole SDLC, and a fractional architect. Roughly 1.5 FTE where a five-person team used to sit. 98% of the code we ship isn’t handwritten.

That’s why engagements shrink: the work gets done, the harness takes more of the load, and we pull people off.

We don’t know how to sell you more headcount.

We grow on demand and on the efficiency we hand each client. That’s the only thing that carries to the next one.

If your vendor’s team on your account has grown since AI arrived, you’re paying for their business model, not yours.

[Limestone Digital @LimestoneHQ]

@mardehaym 👀

[Mark Ajzenstadt @mardehaym]

The difference between an AI consultancy and a hands-on implementation firm is simple.

One sells a long transformation. The other ships work, gets first numbers on the board, and iterates from real metrics. That is the better path: implement, measure, adjust, move.

Not a year-long generic AI program with workshops, frameworks, and no operating data.

Every engagement we run at @LimestoneHQ has fewer people on it now than when it started. The firm is growing faster than it ever has.

A consultancy has to keep an expensive person busy.

$350 an hour, billed by the day. The work is built to last. Scope expands. The team expands. Phase two gets sold in month three. You end up with a $500K deck about “unlocking potential.”

Ask people two or three levels below the executive who signed. They can’t stand consultants. They watched the circus.

What the firm is really selling is cover: I hired a top firm, so I made the responsible choice. If it fails, I still hired a top firm and there's nothing else we can do.

For a mid-market company, that’s an expensive way to avoid deciding.

Consultants do earn it sometimes. A dynamic consultant at $5K an hour who unblocks the thing that’s been strangling operations for two years is worth every dollar. A specific, painful problem for the business solved. That’s a fair trade.

Generic advice is the trap. In 2026, if the advice isn’t tied to a real problem, you can open a model, ask for the top five strategy frameworks, give the output to your HR lead with deliverables attached, and start Monday.

You’ll learn more from the first failed iteration than from three months of workshops. If you need a $350-an-hour outsider to tell you what your strategy is, the honest move is to change your job.

Vendors are selling the same billable-day model with an AI sticker on it. When a mid-sized or large engineering shop says they’ve gone AI-native, I ask one question: how many people have you let go, and why are you still hiring more? If the answer is none and lots, they aren’t passing any efficiency to you.

They added “AI” to the rate card and kept the man-day model underneath.

The small firms are the newest version. A group assembles a team to ride the AI demand wave, hires people with case studies, and suddenly the firm has case studies. But there’s no history behind the logos.

Our @LimestoneHQ AI velocity pod is one senior engineer, agents across the whole SDLC, and a fractional architect. Roughly 1.5 FTE where a five-person team used to sit. 98% of the code we ship isn’t handwritten.

That’s why engagements shrink: the work gets done, the harness takes more of the load, and we pull people off.

We don’t know how to sell you more headcount.

We grow on demand and on the efficiency we hand each client. That’s the only thing that carries to the next one.

If your vendor’s team on your account has grown since AI arrived, you’re paying for their business model, not yours.

[Limestone Digital @LimestoneHQ]

Want help measuring your AI pilot before it starts? Book a discovery call with us:

https://t.co/wW4omvxvc4

[Limestone Digital @LimestoneHQ]

An AI pilot report that says "hours saved" hasn't measured anything a CFO can bank yet.

We measure pilots in 6 steps, agreed before the build starts:

[Limestone Digital @LimestoneHQ]

Want to put AI to work in your operations next?

Get our free guide: how to pick your first workflow, measure the results and scale what works, with 3 case studies from healthcare, insurance and commercial lending.

https://t.co/9NAJfWRb3t

[Limestone Digital @LimestoneHQ]

@mardehaym 👀

[Mark Ajzenstadt @mardehaym]

None of this is satire.

→ Jensen Huang says a $500,000 engineer should burn at least $250,000 a year in AI tokens

→ Meta staff burned 73.7 trillion tokens in a month on an internal leaderboard. Some left agents running idle just to climb it

→ Meta is now putting teams on token budgets

→ Apple reportedly cut engineers' daily AI limit from $300 to $200, with $25 next for some teams

→ Uber capped AI coding spend at $1,500 per engineer per month, after burning its 2026 budget by April

→ Goldman took Claude Code off its bankers' desks

→ Revolut moved 2,000 engineers to a self-hosted model and says the bill dropped 90%

→ A law firm told our team their Claude bill went up 10x when their plan switched to usage pricing

Burning tokens doesn't equal delivering value.

[Limestone Digital @LimestoneHQ]

@mardehaym 👀

[Mark Ajzenstadt @mardehaym]

“We’ll build your AI agent. But the IP stays with us.”

Let’s call this vendor model what it is: a raw deal.

You finance the R&D, your frontline experts spend weeks teaching the vendor your domain, and your proprietary data serves as the training ground.

Then, when the engagement ends, you’re handed a usage license for an asset *you* funded, while the vendor packages your institutional knowledge to sell to your competitors.

To be clear: not every solution requires a custom build. If a SaaS tool solves your problem, buy it. Off-the-shelf software is cost-effective, scalable, and somebody else handles the tech debt.

What we reject is charging custom engineering rates for an asset the client will never own.

The real value in enterprise AI comes from building an agent around how *your* business operates. That advantage can’t be bought off the shelf. It lives in your operational edge cases, in the judgment of the specialist who knows why certain claims fail, and in the custom workflows holding legacy systems together.

Here’s how we approach building enterprise AI:

1. Start at the edge, not the top: Work alongside the people running the process before drafting a single prompt. Go beyond the executive brief and talk to the operational experts who handle the exceptions.

2. Real-world edge cases become the benchmark: Turn those cases into tests. The agent is measured against actual historical cases, and the domain experts define what success looks like.

3. Build inside the client’s environment: Every line of code, orchestration layer, and prompt is built directly inside your cloud environment, under your governance and accounts.

4. Give one person operational accountability: Assign one internal owner who manages the process today, governs agent behavior, and has final sign-off on deployments.

What this looks like in practice:

A PE-backed healthcare billing platform we partnered with relied on a 20-year-old legacy rules engine. No packaged AI vendor on the market could deploy into that environment out of the box.

Within six weeks of kickoff, their CPTO was evaluating a production-ready agent. It scored 59/60 on its initial deployment, runs within their own cloud tenant for $250 a month in total operating costs, and delivered immediate ROI. They expanded our engagement.

An AI agent built from your institutional knowledge, your data, and your capital belongs to your business. Period.

If a vendor wants to keep the IP, they should fund the build themselves.

[Limestone Digital @LimestoneHQ]

Got a risky migration coming up? Book a discovery call with us:

https://t.co/WK8DEzuhgT

[Limestone Digital @LimestoneHQ]

The migration that takes production down is the one somebody tried to ship in a single PR.

The safe way has always been seven small PRs, and teams skipped it because it was tedious. Agents write tedious code all day, so that excuse is gone:

1. Expand. Add the new column or table, nullable, with nothing reading it, and deploy that alone.

2. Dual write. The app writes to old and new behind a flag, and reads stay on old.

3. Backfill. A throttled job copies old rows in small batches, can stop and restart at any row, and only writes rows the new column hasn't already got.

4. Verify. A script compares old and new row by row and lists every mismatch. The count has to reach zero.

5. Switch reads. Move reads behind the flag by percentage, watch errors and latency, and keep dual writes on through a full business cycle, including your month-end close or billing run, so you can flip back.

6. Stop old writes. Delete the old-column code and the flag, and deploy.

7. Contract. Drop the old column in its own PR, once query logs show nothing still reads it.

An agent can write all seven PRs, the backfill job and the comparison script. A person approves steps 5 and 7, and step 7 waits for a backup you've restored from at least once.

Tell the agent two things it can't work out from the repo: which tables lock on ALTER at your row counts, and which reports or nightly jobs outside the codebase read that column.

[Limestone Digital @LimestoneHQ]

Want to put AI to work in your operations next?

Get our free guide: how to pick your first workflow, measure the results and scale what works, with 3 case studies from healthcare, insurance and commercial lending.

https://t.co/9NAJfWRb3t