AI Consulting for Businesses: What You Really Need

AI Consulting for Businesses: What You Really Need

Artificial Intelligence is the buzzword of the hour. Everyone wants AI, everyone offers AI consulting. But what do you really need? This article separates hype from reality.

Table of Contents

1. What Does AI Consulting Actually Do?
2. The Problem with Pure Strategy Consulting for AI
3. AI Readiness: Is Your Company Ready?
4. The 4 Most Common AI Use Cases for Mid-Market
5. Why 70% of AI Projects Fail
6. From AI Strategy to Production System
7. AI Consulting Costs
8. Red Flags in AI Consultants
9. Checklist: Finding the Right AI Consulting

---

What Does AI Consulting Actually Do?

Before we dive in: What should AI consulting deliver?

The Three Pillars of Good AI Consulting

1. Use Case Identification Where can AI create concrete value in your company? Not every task needs AI – and not every AI solution fits every problem.

2. Feasibility Analysis Do you have the data? The infrastructure? The processes? AI is not a magic wand – it needs prerequisites.

3. Implementation And this is where the wheat separates from the chaff: Who can actually deliver?

What AI Consulting is NOT

• Not a tool demo: "Look, ChatGPT can do this" is not consulting
• Not a trend presentation: Slides about AI trends don't help you
• Not strategy without a plan: "You should use AI" is not an insight

---

The Problem with Pure Strategy Consulting for AI

This gets uncomfortable. Most AI consultations fail due to a fundamental problem:

The Theory-Practice Gap

What traditional consultants deliver:

• Market analyses
• Use case prioritization
• Roadmaps
• Governance frameworks
• Beautiful presentations

What they DON'T deliver:

• Working systems
• Integrated workflows
• Trained models
• Productive solutions

A Typical Scenario

Phase 1: The Strategy (3-6 months, €150,000)

• Stakeholder interviews
• Process analysis
• Use case workshops
• Roadmap development

Result: A document with "Top 10 AI Use Cases" and a 3-year roadmap.

Phase 2: The Disillusionment

• Consultants leave
• IT department overloaded
• No internal AI know-how
• Search for implementation partner
• Budget exhausted

Result after 18 months: Not a single AI system in production.

The Fundamental Problem

AI is not a strategy discipline. AI is an engineering discipline.

You cannot advise on AI without being able to build it. That would be like an architect designing bridges without understanding structural engineering.

---

AI Readiness: Is Your Company Ready?

Before investing in AI, you should honestly check: Are the prerequisites met?

The AI Readiness Check

1. Data Maturity

Level	Description	AI Suitability
1	Data scattered in Excel, emails, folders	❌ Not ready
2	Central database, but no quality assurance	⚠️ Conditional
3	Structured data with quality standards	✅ Ready
4	Data pipelines, automatic validation	✅ Very good

2. Process Maturity

Level	Description	AI Suitability
1	Processes not documented	❌ Not ready
2	Processes documented but not standardized	⚠️ Conditional
3	Standardized, measurable processes	✅ Ready
4	Automated processes with KPIs	✅ Very good

3. Technical Infrastructure

Level	Description	AI Suitability
1	Legacy systems, no APIs	❌ Not ready
2	Modern systems, but isolated	⚠️ Conditional
3	Integrated system landscape	✅ Ready
4	Cloud-native, API-first	✅ Very good

4. Organizational Maturity

Level	Description	AI Suitability
1	"We've always done it this way"	❌ Not ready
2	Openness, but no capacity	⚠️ Conditional
3	Dedicated team or budget	✅ Ready
4	Innovation culture, agile working	✅ Very good

Honest Assessment

If you're below Level 2 in more than one category: Don't start with AI.

Start with:

• Data consolidation
• Process standardization
• System integration

This sounds less sexy than "AI transformation" but delivers more.

---

The 4 Most Common AI Use Cases for Mid-Market

1. Intelligent Document Processing

The Problem: Incoming documents (invoices, contracts, inquiries) are manually read, categorized, and processed.

The AI Solution:

• Automatic classification
• Data extraction (OCR + NLP)
• Structured handoff to downstream systems

Typical ROI:

• 70-90% time savings
• Error reduction from 8% to <1%
• Break-even in 4-8 months

Complexity: Medium Prerequisites: Digital documents, structured target systems

2. Customer Service Automation

The Problem: Support team answers repetitive questions, long wait times, inconsistent responses.

The AI Solution:

• Chatbot for first-level support
• Automatic ticket categorization
• Response suggestions for agents

Typical ROI:

• 40-60% of inquiries automated
• Response time from hours to seconds
• Employee satisfaction increases (less routine)

Complexity: Medium to High Prerequisites: Knowledge base, historical ticket data

3. Predictive Analytics (Sales & Demand)

The Problem: Sales works leads by gut feeling. Orders are planned reactively.

The AI Solution:

• Lead scoring based on behavioral data
• Demand forecasting for inventory planning
• Churn prediction for customer retention

Typical ROI:

• Win rate +15-25%
• Inventory costs -10-20%
• Customer churn -20%

Complexity: High Prerequisites: Historical data (min. 2 years), CRM/ERP integration

4. Process Automation with AI

The Problem: Workflows require human decisions that are actually rule-based.

The AI Solution:

• Intelligent workflow control
• Automatic approvals with anomaly detection
• Dynamic prioritization

Typical ROI:

• Throughput time -50-70%
• Capacity gain 2-3 FTEs
• Compliance improved (complete documentation)

Complexity: Medium Prerequisites: Defined processes, digital workflows

---

Why 70% of AI Projects Fail Without Operational Implementation

The numbers are sobering: According to Gartner, 70% of AI projects never reach production. Why?

Reason 1: POC Trap

The Pattern:

1. Consultants build impressive Proof of Concept
2. Demo works wonderfully with test data
3. Project "successfully" completed
4. POC gathers dust because:
   • Integration too complex
   • Data quality poor in real system
   • No operations concept
   • No accountability

The Lesson:

A POC is not a product. It only proves something is technically possible – not that it works practically.

Reason 2: Data Reality

In the workshop: "We have all the data, we just need to combine it."

In reality:

• Data in 7 different formats
• 30% missing values
• Inconsistent naming
• No historical depth
• GDPR issues

The Lesson:

80% of an AI project is data work. Underestimate this and you fail.

Reason 3: The Integration Problem

AI models are useless if they're not integrated into existing processes.

Example:

• You have a model that classifies customer inquiries
• It achieves 95% accuracy
• But: How does the inquiry come in? How does it go out? Who monitors errors?

The Lesson:

AI without integration is a toy. Integration is often more complex than the model.

Reason 4: Organizational Resistance

Typical Reactions:

• "We can do that better"
• "I don't trust that"
• "Who's responsible if the AI is wrong?"
• "That's taking our jobs"

The Lesson:

Change management is not nice-to-have. Without acceptance, no usage.

Reason 5: The Maintenance Blind Spot

AI systems are not install-and-forget solutions.

What happens after go-live:

• Models degrade (data drift)
• Edge cases emerge
• Requirements change
• Updates needed

The Lesson:

Plan for 20-30% of initial budget for annual maintenance.

---

From AI Strategy to Production System

How do you do it right? Here's a pragmatic approach:

Phase 1: Quick Assessment (1-2 weeks)

Activities:

• Understand existing processes
• Map data landscape
• Identify quick wins
• Check feasibility

Result:

• 3-5 concrete use cases with assessment
• Data gap analysis
• Recommendation for first pilot

Cost: €3,000-8,000

Phase 2: Focused Pilot (4-8 weeks)

Activities:

• Fully implement ONE use case
• Work with real data
• Integration with existing systems
• User training

Result:

• Productive AI system
• Measurable results
• Lessons learned for scaling

Cost: €15,000-50,000

Phase 3: Scale & Optimize (ongoing)

Activities:

• Implement additional use cases
• Optimize existing solutions
• Monitoring and maintenance
• Knowledge transfer

Cost: Variable, typically €5,000-15,000/month

The Difference from the Traditional Approach

Traditional	Pragmatic
6 months strategy, then RFP	2 weeks assessment, then pilot
€200,000 for concept	€50,000 for productive system
18 months to first result	8 weeks to first result
High expectations, often disappointment	Quick wins, iterative improvement

---

AI Consulting Costs: What's Realistic?

Understanding Cost Structure

Initial Costs:

Component	Budget Range
Assessment/Strategy	€3,000-30,000
Pilot/POC	€15,000-80,000
Production system (simple)	€30,000-100,000
Production system (complex)	€100,000-500,000

Ongoing Costs:

Component	Budget Range/Year
Cloud/Infrastructure	€2,000-20,000
API costs (OpenAI, etc.)	€500-10,000
Maintenance & Optimization	€5,000-30,000
Support	€3,000-15,000

Market Day Rates

Consultant Profile	Day Rate
Junior AI Consultant	€800-1,200
Senior AI Consultant	€1,200-1,800
AI Architect/Lead	€1,800-2,500
Big 4 / McKinsey	€2,500-5,000
Freelance ML Engineer	€800-1,500

ROI Orientation

A sensible AI investment should pay off within 12-18 months.

Rules of Thumb:

• Document processing: ROI in 4-8 months
• Customer service bot: ROI in 6-12 months
• Predictive analytics: ROI in 12-18 months
• Complex automation: ROI in 12-24 months

---

Red Flags in AI Consultants

🚩 Red Flag 1: Strategy Only, No Implementation

Warning Sign: "We develop the AI strategy, for implementation we recommend partners."

Problem: Those who can't implement don't understand feasibility.

🚩 Red Flag 2: Buzzword Bingo

Warning Signs:

• "Transformative AI journey"
• "AI-first mindset"
• "Cognitive enterprise"
• More buzzwords than concrete solutions

Problem: Empty words hide lack of substance.

🚩 Red Flag 3: No References in Your Industry

Warning Sign: "We have lots of AI experience" – but no concrete cases in mid-market.

Problem: AI for a corporation is completely different from AI for a 200-person company.

🚩 Red Flag 4: Unrealistic Promises

Warning Signs:

• "100% automation possible"
• "Productive in 4 weeks"
• "The system learns by itself"

Problem: Those who promise miracles deliver disappointment.

🚩 Red Flag 5: No Interest in Your Data

Warning Sign: Proposal without questions about data quality and availability.

Problem: Those who don't understand the data situation can't assess feasibility.

🚩 Red Flag 6: Vendor Lock-in

Warning Signs:

• Proprietary platform without export
• No access to trained models
• Long-term contract required

Problem: Dependency on one provider is a risk.

---

Checklist: Finding the Right AI Consulting

Before the Meeting

• Own pain points identified
• Rough idea of the goal
• Budget range defined
• Stakeholders identified

Check in Initial Meeting

• Does the consultant ask about your data?
• Do they show interest in your processes?
• Do they name concrete, similar projects?
• Do they also explain what DOESN'T work?
• Do they offer a small first step?

Check Proposal

• Clear deliverables defined?
• Milestones with measurable results?
• Transparent pricing structure?
• Exit options if unsuccessful?
• Maintenance and support included?

Check References

• Concrete case studies requested
• Spoke with reference customers
• Results quantified (not just "successful")
• Similar company size/industry

After Project Start

• Regular updates agreed
• Access to all work results
• Knowledge transfer planned
• Success metrics defined

---

Conclusion

AI consulting can transform your business – or burn a lot of money. The difference lies in choosing the right partner.

Key Takeaways:

1. Strategy without implementation is worthless. Choose partners who can do both.

2. Start small. A successful pilot is worth more than a perfect roadmap.

3. Data is key. Without good data, no good AI.

4. Be skeptical of promises. If it sounds too good, it usually is.

5. Plan long-term. AI is not a one-time investment, but a journey.

The best AI projects are not the most ambitious – but those that actually go into production.

---

Looking for AI consulting that doesn't just advise but implements? Let's talk – no obligation, concrete results.