Pillar · AI Strategy

AI Strategy for Service Providers: Diagnose Before You Deploy

Klarna's CEO promised AI would do the work of 700 humans. Fifteen months later, he was hiring humans back. Service providers about to make the same bet should run the diagnostic first.

By Colin Taylor · Founder, Asset Alchemy · Wake Forest, NC

Last updated · June 7, 2026  ·  12 min read

On This Page
01The Twelve Months That Cost Klarna $99 Million 02Why Most Service Providers Are One Decision Away From the Klarna Mistake 03Three Substitutes That Lead to the Same Ending 04The Two Diagnostics That Would Have Saved Klarna $99 Million 05The Twelve Months Klarna Needed, and Skipped 06Five Signs Your Business Is Already on the Klarna Path 07Four Service Businesses That Found the Path Klarna Didn't 08What the Calendar Looks Like When the System Holds Without You 09Frequently Asked Questions 10About the Author

This morning, somewhere, a successful service business is about to sign an AI contract that will quietly hollow it out.

Not catastrophically. Not all at once. Just enough that twelve months from now, the founder will look at their pipeline, their content output, their team capacity, and realize the tools they bought to amplify the business have instead been competing with it.

The diagnostic that prevents this takes 45 minutes. Most won't run it. That is exactly why most AI strategies will fail.

01 · The Story

The Twelve Months That Cost Klarna $99 Million

In February 2024, Sebastian Siemiatkowski stood up and announced that Klarna's new AI chatbot was doing the work of 700 humans. Customer service handle time had dropped from 11 minutes to under 2. Forty million dollars in projected "profit improvement." Two-point-three million customer conversations handled. Silicon Valley applauded. TechCrunch ran the cover. The press release went out under the headline you would expect.

Fifteen months later, the same CEO stood up and admitted on the record that the all-AI strategy had produced "lower quality" customer service. Klarna would resume hiring human agents.

What changed was not the AI. The AI worked. It cut costs exactly as promised. What collapsed was everything the AI had been deployed on top of: the institutional knowledge, the customer trust, the texture of judgment that experienced service representatives brought to complex problems. None of that had been extracted before the automation went in. So when the chatbot handled the easy two-thirds of conversations beautifully and stumbled on the hard remainder, there was no documented capability left to absorb the impact. Customer dissatisfaction surged. Complex issues piled up. The headcount that had been celebrated as savings became a quality problem the company eventually had to reverse in public.

Siemiatkowski's own words, said quietly in private rooms with other Silicon Valley CEOs and later admitted publicly: "I talk to them in private rooms and it's like, 'Oh my God, the jobs are gone.' I feel that is dishonest."

This is the pattern. The technology performs. The implementation succeeds on its own terms. The business loses anyway, because the foundation underneath was never built, and the AI made the gap visible faster than anyone expected.

"AI does not fail because the technology is flawed. AI fails because businesses deploy it on top of undocumented expertise, unextracted knowledge, and founder-dependent operations."

Klarna had public-markets ambitions, the best AI engineering talent in fintech, and OpenAI as a direct partner. They still got it wrong. The question is not whether your business is too small to make this mistake. The question is whether the mistake is happening to you right now, in slower motion, on a smaller scale, with consequences that show up as missed quarters rather than headline-grade reversals.

02 · The Problem

Why Most Service Providers Are One Decision Away From the Klarna Mistake

Klarna's collapse is unusual only in scale. The pattern that produced it (capability deployed before capability was documented) operates in every service business that pursues AI without first running the diagnostic underneath it.

The AI strategy conversation has been dominated by product companies, SaaS platforms, and enterprise IT departments. Their playbook is straightforward: identify a process, select a tool, integrate it, measure output. The tools are good. The playbooks are well-tested. The case studies are real.

None of it applies to a service business.

Your product is your expertise. Your competitive advantage is institutional knowledge. Your revenue depends on relationships and trust that no platform can replicate. When a consultant, coach, financial advisor, or professional firm tries to apply a product-company AI strategy, the results are predictable: expensive tool subscriptions that nobody uses, automations that strip out the human judgment clients are paying for, and content engines that flood channels with generic output that actively erodes the authority those professionals spent years building.

The numbers underneath this are stark. IBM's 2025 CEO Study, which surveyed 2,000 chief executives across 33 countries, found that only 25% of AI initiatives have delivered the expected ROI. MIT's NANDA Initiative reported in 2025 that 95% of generative AI pilots failed to show measurable impact on the bottom line. Forrester Research's Predictions 2026: The Future of Work found that 55% of companies regret the workforce reductions they made in anticipation of AI capabilities. The Cloudflare outage of November 2025 made the dependency problem visible in a different way: the businesses that survived the disruption best were the ones with documented internal capabilities. The ones that struggled most were the ones who had outsourced their thinking to tools they did not control.

The service provider who has spent fifteen years developing a methodology for transforming client outcomes has something infinitely more valuable than any AI tool: proven, differentiated expertise. The problem is that this expertise has never been extracted, documented, or structured in a way that anything (human or artificial) can reliably access and amplify. This is the diagnostic gap. It is the single largest reason AI strategies fail for service providers. Not because the AI was wrong. Because the foundation was never built.

03 · The Misdirection

Three Substitutes That Lead to the Same Ending

The three approaches most service providers reach for when they decide to "get serious about AI" all fail for the same reason. Each one tries to add capacity without first making the underlying capability transferable.

Hire more. The instinct is to expand the team so the founder is less of a bottleneck. But every new hire inherits the same broken handoff: institutional knowledge still lives in the founder's head, decisions still route through the founder's calendar, and the new person produces work that has to be re-done because the standard was never written down. Adding people to an undocumented operation multiplies the chaos rather than reducing it. This is the dynamic I've called the Competence Trap: the better the founder is, the harder the business is to scale.

Automate more. Workflow tools, scheduling automation, CRM sequences, marketing platforms. Each individually useful; collectively a small disaster. Automation that runs on top of an undocumented process just runs the chaos faster. You end up with a beautifully sequenced system that produces work no one is sure how to evaluate, because the criteria for evaluation never got written down either. The 2026 AI Gut-Check walks through what this looks like in practice.

Buy AI tools. The newest version of the same impulse. Lovable, Cursor, ChatGPT for Teams, an enterprise Copilot license, a vertical-specific RAG system. The tools work. The output is real. But the output reflects the inputs, and if the inputs are an undocumented founder's brain, the AI amplifies that vagueness at scale, producing content and decisions that look like the founder's work but lack the judgment that made the original valuable. There is a name for this: AI dependency, the inverse of AI amplification. Dependency is what you get when AI replaces a foundation that never existed. Amplification is what you get when AI builds on a foundation that does.

The unifying problem with all three approaches is that they treat the AI question as a tooling question. It is not. It is a documentation question. And until the documentation exists, no tool, no hire, and no automation will produce sustained gain.

04 · The Mechanism

The Two Diagnostics That Would Have Saved Klarna $99 Million

Klarna had access to the best AI engineering talent fintech could buy. Board oversight. Capital. OpenAI as a direct partner. Twelve full months of runway to course-correct. None of it mattered, because two specific diagnostic frameworks were missing from their pre-implementation work. Both can be run today by any service provider, to see the threat profile before it materializes and the asset base before it gets automated away.

The first describes the threats. The second describes the assets. Neither involves a tool recommendation. Both are required.

The D.I.B.S. Dilemma: four forces eroding your business value

The four market forces compounding against service businesses right now. Each one individually solvable. Together, they create a threat profile no single AI tool addresses.

Decision Fatigue. Buyers are overwhelmed. Every week brings a new platform, a new vendor, a new promise. The cognitive cost of evaluating options has become so heavy that buyers default to the safest choice: doing nothing. The diagnostic defense is that when your methodology is documented and your results are structured as proof assets, buyers do not need to evaluate you against every other option. Your framework becomes the decision shortcut.

Inflationary Pressures. AI implementation costs are climbing, not falling. Integration, training, customization, and maintenance routinely push total costs to 3x to 5x the quoted price. The diagnostic defense is that the Asset Alchemy diagnostic consistently identifies $20,000 to $50,000 in dormant revenue from existing assets, revenue that can fund AI implementation from a position of strength rather than scarcity.

Buyer Bottlenecks. 86% of purchases now stall before completion. AI has made it worse. Every channel is flooded with competing claims and synthetic testimonials, and buyers cannot tell who is real. So they stall. The diagnostic defense is that documented methodologies and structured proof assets cut through synthetic noise because they carry specificity and depth that AI-generated content cannot replicate.

Synthetic Content. AI-generated content now dominates LinkedIn, email, and search results. Every channel that once built trust now actively erodes it as buyers develop resistance to anything that feels manufactured. The diagnostic defense is content derived from documented institutional knowledge, the kind of writing that carries the fingerprints of real experience and produces insights generic AI cannot.

The K.A.S.H. Framework: four categories of extractable business value

Every established service provider has these four categories of assets inside their operation. The difference between a business worth its revenue and a business worth multiples of its revenue is whether these have been extracted and structured for activation. The CLEAR Protocol describes the operating discipline that makes this extraction possible without disrupting client delivery.

K, for Knowledge. The institutional expertise trapped in the founder's head. Industry insights developed over decades, pattern recognition from hundreds of engagements, diagnostic intuition that identifies problems others miss. When extracted, knowledge becomes training material, content frameworks, diagnostic tools, and AI-powered knowledge bases that serve clients at scale.

A, for Assets. Content, databases, client relationships, intellectual property, and resources already created. Most service providers have hundreds of pieces scattered across drives, inboxes, and memory. When extracted, assets become organized libraries, repurposable content systems, and relationship activation campaigns that generate revenue from dormant connections.

S, for Systems. Documented processes, playbooks, and operational workflows. For most service providers, these exist as habits in the founder's routine rather than documented procedures anyone else could follow. When extracted, systems become transferable operating procedures, delegation frameworks, and the infrastructure AI tools actually need to function.

H, for Habits. Decision patterns, client interaction frameworks, and delivery rhythms that produce consistent results. These unconscious competencies separate expert practitioners from generalists. When extracted, habits become your Signature Method, the documented delivery framework that is the single most valuable asset in any service business.

05 · The Method

The Twelve Months Klarna Needed, and Skipped

If Klarna had run the diagnostic before the AI contract was signed, the twelve months that followed would have looked nothing like the ones they lived through. The work happens in three phases across nine steps. Each step produces a deliverable the client owns permanently. No step is theoretical. No step requires a tool purchase before the underlying capability is documented.

Phase 1, Clarity (Steps 1 through 4). The Asset X-Ray is the complete diagnostic inventory: every hidden asset identified, extraction readiness scored, dormant revenue mapped. The Resource Optimizer audits where time, money, and energy are being wasted and reclaims that capacity before new tools are added to the stack. The Market Advantage Map identifies the Category-of-One position and the defensible moats that survive AI commoditization. Buyer Desires closes the phase with a buyer-psychology blueprint that maps the confidence gap between what buyers need to believe and what the current proof architecture actually demonstrates.

Phase 2, Confidence (Steps 5 and 6). Oxygen Offers rebuilds the offer architecture on documented assets, producing offers that cannot be commoditized because they are built on a proprietary diagnostic framework. Signature Method is the methodology extraction itself, the most valuable deliverable in the system. The delivery methodology captured as a named, structured, defensible framework the client owns permanently.

Phase 3, Control (Steps 7 through 9). The Revenue Engine builds infrastructure that converts documented assets into predictable revenue independent of the founder's calendar. The Cashflow Catalyst activates documented assets in the market, which is where the $20K to $50K in dormant revenue typically materializes. Brand Boomerang uses documented methodology and structured content to build authority that compounds and attracts clients instead of chasing them.

The sequencing matters. Skipping ahead to revenue activation without first completing the extraction produces brittle gains that depend on the founder's continued involvement. The method does not assume AI is needed. It assumes clarity about what you already have is needed first.

06 · The Signs

Five Signs Your Business Is Already on the Klarna Path

01
You are the documentation
If a new hire asks how a client engagement starts, the honest answer is "ask me." If a team member needs to know what makes a proposal successful, they shadow you. If the answer to most "how do we…" questions lives in your head and nowhere else, the AI you deploy will inherit the gap, not close it.
02
Your team produces work you have to fix
Output technically exists. It is wrong in ways you can identify but cannot articulate quickly enough to teach. The criteria for "good" have never been written down because they have never needed to be. Every piece of output requires you to be the quality filter, which means the team's capacity is capped at your review bandwidth.
03
You cannot take a vacation without checking in
Two weeks away surfaces problems no one else can resolve. Slack messages accumulate. Decisions wait. The team is competent at execution but dependent on you for judgment calls that have never been codified. The business runs on your attention, not your systems.
04
Your AI experiments produce generic output
You have tried ChatGPT, Claude, vertical AI tools. The output reads like it could have come from anyone in your category. That is the signal that the AI has nothing proprietary to draw on. You are feeding it the same public training data your competitors are. The differentiation lives in your undocumented judgment, which the AI cannot access.
05
Your content engine has stalled or turned synthetic
You stopped publishing because you ran out of time. Or you kept publishing but the work no longer sounds like you. Both symptoms point to the same root: the institutional knowledge that fueled the original content was never extracted, so the moment you scaled, the source dried up.

Three or more of these symptoms in a current operation is the signal the foundation work is overdue. The four businesses below recognized them in time and acted differently.

07 · The Evidence

Four Service Businesses That Found the Path Klarna Didn't

Retirement Advisor · $4M New AUM
72 content assets extracted from a few hours of recorded conversation
A retirement advisor with twenty years of proprietary methodology had nothing documented. The Asset X-Ray and Signature Method extraction produced 72 structured content assets from existing client recordings. Those assets became the basis for a thought-leadership engine, a referral system, and the AUM growth that followed. The methodology had always existed. It had just never been visible.
Culture Coach · $250K+ in multi-year contracts
Invisible to fully booked, 40+ hours per month reclaimed
A culture coach with strong delivery and no consistent pipeline went from "invisible to the market" to a fully booked calendar with multiple multi-year engagements. The lever was not new marketing. It was a documented offer architecture built on extracted IP. Once buyers could see the methodology, the decision became easy. Forty-plus hours per month reclaimed from work that had previously required the founder personally.
Healthcare CEO · $20K saved, 7 days
Automated workflows that run without the founder
A healthcare CEO had built a thriving practice but could not step back without operations degrading. The Resource Optimizer and Revenue Engine work documented the operational habits, then automated the routine layer. The result: $20K in annualized savings within seven days, and a business that continues to operate when the CEO is unavailable.
Hidden Asset case · $37K+ from a single voice note
A 30-minute cleanup that produced a five-figure offer
A consultant recorded a thirty-minute voice note while driving, stream-of-consciousness on a client problem. Standard treatment of such a note is to file it and forget it. Structured extraction turned the same recording into a documented methodology, a positioning insight, and an offer that produced $37K in revenue within the engagement window. The asset had existed for less than an hour. The extraction made it visible.
08 · The After-State

What the Calendar Looks Like When the System Holds Without You

Imagine taking three weeks off and returning to find the business stronger than when you left. The team executed against documented standards rather than waiting for your judgment. The content engine ran on extracted methodology rather than your weekend writing time. The pipeline filled with prospects who arrived already understanding the framework, asking informed questions, ready to discuss engagement rather than to be educated.

Imagine evaluating a new AI tool and being able to answer in a single conversation whether it amplifies your existing assets or merely adds another subscription. The diagnostic gives you a structured map of what the business is, which makes every subsequent tooling decision faster, cheaper, and more defensible. You stop chasing platforms because you know what you are looking for. You stop second-guessing implementations because you have a foundation against which to measure them.

Imagine the conversations with prospects shifting in tone. The work of proving you are worth a premium ends, because the documented methodology does that proving on its own. The proposal becomes a confirmation rather than a sales document. The pricing power moves up. The pipeline shortens. The wins compound.

None of this requires new technology. All of it requires the foundation to be built first.

09 · Questions

Frequently Asked Questions

What is a diagnostic-first AI strategy?
Auditing and documenting your existing business assets before selecting any AI tools. The Asset Alchemy Method uses a nine-step diagnostic to extract and score assets before any recommendation is made. The sequence matters because AI deployed on undocumented operations amplifies the gaps rather than closing them.
Why do most AI strategies fail for service providers?
They start with the tool instead of the foundation. AI layered on undocumented processes automates chaos rather than amplifying value. Industry research consistently shows AI project failure rates in the 70 to 80 percent range; the common factor is missing diagnostic groundwork, not flawed technology.
What is the D.I.B.S. Dilemma?
Four market forces eroding service business value: Decision Fatigue, Inflationary Pressures, Buyer Bottlenecks, and Synthetic Content. Each one is individually solvable; together they create a compounding threat that no single AI tool addresses. The diagnostic surfaces which of the four is hitting hardest in a specific business.
What is the K.A.S.H. framework?
Knowledge, Assets, Systems, and Habits. The four categories of extractable value inside every established service provider. Knowledge is the expertise in the founder's head. Assets are content and relationships already created. Systems are processes that exist but have not been documented. Habits are the unconscious competencies that produce consistent results. The method extracts and activates all four.
How long does it take to build an AI-ready foundation?
Ninety days. The Foundation Sprint runs the nine-step diagnostic across three phases: Clarity, Confidence, and Control. Most clients activate $20,000 to $50,000 in dormant revenue during the process itself, which means the engagement frequently funds itself before completion.
Can I use AI tools alongside the diagnostic?
Yes, strategically. The diagnostic identifies which tools amplify your assets and which add complexity without return. Many clients continue using their existing AI subscriptions during the engagement; what changes is the discipline around which tools are kept, which are paused, and which need to wait until the underlying capability is documented.
What size business does this work for?
Established service providers at $500,000 to $10M+ in annual revenue. Consultants, coaches, financial advisors, healthcare practitioners, agency owners, and B2B service firms with real expertise that has never been systematically extracted. Below that range the documentation work is usually premature; well above that range it tends to be too late to do without dedicated transformation resources.
What if I have already invested in AI tools that are not delivering?
That is one of the most common starting points. The diagnostic audits existing tools as part of the Resource Optimizer phase and identifies which subscriptions are amplifying value, which are dormant, and which are actively competing with the business. Recovering wasted tool spend typically funds a meaningful portion of the engagement before any new revenue activation begins.
Do I need a CTO or technical team to do this?
No. The method does not require coding, custom integrations, or a technical lead. The work is done in strategic conversations and documented in formats the founder owns and can use directly. The technology layer is handled on the Asset Alchemy side. Clients bring expertise and time. The implementation is the engagement's job.
How is this different from buying ChatGPT for Teams or a vertical AI platform?
Those are tools. The diagnostic-first approach is the foundation those tools need to work. ChatGPT or a vertical platform deployed on top of extracted methodology produces dramatically better output than the same tool deployed on top of an undocumented operation. The two approaches are not in competition. One is the prerequisite for the other.
What does the engagement actually involve week to week?
Focused strategic conversations rather than homework assignments. Most clients describe the time commitment as the most productive few hours of their month. AI-powered extraction tools handle the heavy documentation work between sessions. The founder brings the expertise. The engagement brings the technology and the diagnostic judgment.
10 · About
CTColin Taylor
Colin Taylor
Founder, Asset Alchemy · Wake Forest, NC

Colin Taylor is the founder of Asset Alchemy and the creator of the Asset Alchemy Method, the nine-step diagnostic system used by service providers, consultants, and professional firms to extract trapped institutional knowledge before deploying AI tools on top of it.

His background is unusual for a business strategist. A former U.S. Navy Search and Rescue Swimmer, he spent the early part of his career training for emergency rescues in the open ocean. That work produced the operational philosophy now underneath the Method: diagnose first, then act. He has spent the twenty years since in digital agency leadership and as a former Apple Business Consultant, working with founders and operators across professional services, healthcare, finance, and B2B technology.

He writes the Asset Alchemy Weekly, publishes long-form analysis at LinkedIn, and works directly with a small number of accomplished service providers each year through Asset Alchemy's private engagement programs.

Stop Building on an Undocumented Foundation

The diagnostic call takes 45 minutes. It maps your hidden assets, scores your D.I.B.S. exposure, and shows you where the $20K to $50K in dormant revenue is hiding, before any AI tool is recommended or deployed.

Book a Diagnostic Call →