AI agency, inhouse or IT consulting - what really pays off?
Short: Three paths to AI marketing automation. A specialized agency is the fastest way with predictable costs - ideal for SMBs with clear use cases. Inhouse build-up only pays off from 100 employees and a long-term AI strategy. Classical IT consulting often delivers concepts without running systems. The answer depends on size, maturity and strategy - not on fashion.
Table of contents 9 sections
- 01The three options at a glance
- 02Qualitative comparison of the three models
- 03Decision matrix by company size
- 04How to choose a specialized AI agency
- 05Common mistake: consulting as substitute for execution
- 06The dual location as additional factor
- 07Hybrid models in practice
- 08Frequently asked questions on model choice
- 09Bottom line: the decision comes before the research
Three paths to AI marketing automation. Each has its moment, each has its price, each has its weaknesses. Choose the wrong one and you burn money on an expensive inhouse build-up or end up with an 80-page consulting concept PDF without a single running system.
At Nordsteg we have been working with automated marketing systems in client projects since 2018, and we regularly support Styrian, Carinthian and Viennese companies on exactly this model decision. Some become our clients, some go inhouse, some stay with classical IT consulting. The honest truth: all three paths have merit - but not in every situation. This article shows the pros and cons of each model, with qualitative comparison and decision matrix.
The three options at a glance
Option 1: Specialized AI agency
What they do: The agency builds the workflows productively, runs them ongoing, adjusts them. The contract specifies measurable outcomes, not just concepts. At Nordsteg we work this way - from the first workshop to ongoing operations.
What you get: A productive system with clearly documented workflows, monthly maintenance, adjustments on business changes, access to experience from many comparable projects.
What you do not get: Complete internal knowledge on every detail aspect. The operational running sits with the vendor.
Option 2: Inhouse build-up
What you do: Set up your own role for AI marketing automation, build competency step by step, possibly supported by external consultants.
What you get: Deep process understanding in your own house, full control over the roadmap, knowledge stays with employees.
What you do not get: Fast results. Recruiting for the role takes 4 to 9 months, onboarding another 3 to 6 months. Until the inhouse team delivers productively, often a year passes.
Option 3: Classical IT consulting
What they do: Consulting analyzes status, recommends tools, creates concepts and roadmaps. Some consultancies also accompany implementation, many do not.
What you get: A structured concept with clear recommendations, tool comparison based on your requirements, roadmap for execution.
What you do not get: A running system. Consulting usually ends with the concept paper. You have to organize execution yourself - with inhouse or agency.
Qualitative comparison of the three models
Instead of juggling fictional concrete prices, a qualitative comparison along seven relevant criteria.
Time-to-value
Agency: First productive workflows in 2 to 6 weeks. The agency has the setup pattern from many projects and can adapt standard workflows quickly.
Inhouse: Realistically 12+ months until the first productive workflow. Recruiting, onboarding, tool selection, build-up, testing - everything has to be developed first.
IT consulting: Concept in weeks, then execution still takes the inhouse or agency phase.
Cost structure
Agency: Monthly retainer plus setup. Predictable, scales with demand, can be paused or shrunk.
Inhouse: High fixed costs from a full-time role. Ancillary costs, hardware, tools, training - all running, even in phases without active demand.
IT consulting: Day rates in the high four-figure range. Per consulting day typically 1,500 to 3,000 euros. That adds up fast without a running system at the end.
Knowledge retention
Agency: Knowledge sits partly external. What you can compensate: demand regular trainings, demand documentation of all workflows, regularly secure source code or workflow exports.
Inhouse: Knowledge stays fully in the house. Risk: if the responsible person leaves, the knowledge often goes with them. Deputy and good documentation are mandatory.
IT consulting: Concept knowledge stays with you. The running implementation knowledge only emerges afterwards - with you or whoever does execution.
Scaling
Agency: Very flexible. More workflows? The vendor allocates more hours. Less demand? The contract is adjusted.
Inhouse: Limited by personnel capacity. Once the one inhouse role is full, you need a second - with all recruiting costs. Scaling takes time.
IT consulting: Concept scaling is possible, but operational execution has to happen separately.
Risk
Agency: Dependency risk. If the agency cancels or works poorly, the workflow stalls. Mitigation: daily-cancelable contracts, agree on source code ownership, demand workflow documentation.
Inhouse: Personnel risk. If the person leaves, the system fails. Mitigation: deputy, documentation, knowledge management.
IT consulting: Concept gap. Recommendations that are never executed cost money without effect. Mitigation: only commission consulting when execution is already planned.
For the broader ROI framework on marketing automation, see our assessment of when marketing automation pays off for SMBs.
Decision matrix by company size
Which model fits which size and maturity? Pragmatic orientation from practice.
SMBs with 10 to 50 employees
Recommendation: specialized agency. Inhouse build-up is usually oversized at this size. An own AI role does not pay off because the operational AI tasks do not fill a full-time position. IT consulting is unnecessary - the scope is small enough that you can implement directly.
When to do otherwise: When AI becomes the core business of the company (e.g. an AI consulting vendor itself), inhouse build-up naturally.
Mid-market 50 to 200 employees
Recommendation: mix of agency and external consulting. Here a short IT consulting phase can make sense to sharpen strategy. Then execution by a specialized agency. First inhouse considerations from 150 employees onwards, when clear strategic significance is recognizable.
When to do otherwise: Anyone with a lot of data competence already in the house (typical for e-commerce or producing companies with own IT) can build inhouse earlier.
Mid-market and larger from 200 employees
Recommendation: inhouse core plus selective external support. At this size an own role pays off. The inhouse person coordinates external specialists for special projects. Hybrid model is often the most productive variant here.
When to do otherwise: Anyone with centralized IT structures as a corporation can use a central data role instead of inhouse marketing IT.
Corporation or large enterprise
Recommendation: full inhouse with central data strategy. Selective external consulting for special questions, but the operational substance stays in the house. This recommendation is not relevant for SMB readers, mentioned for completeness.
How to choose a specialized AI agency
If the decision falls on agency, the second question comes: which agency. Six criteria that in our experience separate the wheat from the chaff.
Criterion 1: demonstrable practical projects
What to check: Can you talk to three existing clients similar in size to your company? Does the agency have case studies with concrete numbers and time frames, not just logos? Are the workflows live or at least demo-visible?
Warning sign: "We have just realigned our strategy, so no references."
Criterion 2: transparent communication
What to check: Does the agency also tell you about failed projects or learnings? Are there honest statements on when AI is not the right lever? Are weaknesses of the own tools named?
Warning sign: Everything sounds like success stories, no project ever had problems. Such agencies either learn nothing or do not tell everything.
Criterion 3: tool stack fits
What to check: Which tools does the agency use? Are they compatible with your existing infrastructure? Do you already have HubSpot - can the agency handle it or do they want to switch to a different stack?
Warning sign: Agency recommends the same tool stack regardless of your requirements.
Criterion 4: contract model
What to check: Are contracts daily-cancelable or are there long ties? What happens to workflows and data at contract end? Who owns the setup?
Warning sign: Minimum term 24 months, workflows are "owned by the agency".
Criterion 5: maintenance model
What to check: How is ongoing maintenance arranged? Fixed hour budget or time-and-materials? Who reacts when a workflow breaks at night? Which response times are committed?
Warning sign: "Maintenance we do as needed." That means in practice: late and expensive.
Criterion 6: cultural fit
What to check: In the first conversation - does the agency immediately come with tool demos or do they ask about your strategy? Are they listening or selling? Can you imagine working with these people for 12 months?
Warning sign: A standard pitch is delivered right at the first conversation without reference to your situation.
Common mistake: consulting as substitute for execution
An anti-pattern we see regularly. Leadership commissions a consultancy, gets an 80-page PDF with recommendations after 3 months, archives it, the daily routine rolls on. Six months later the company is again looking for a solution.
What happened: Consulting was confused with execution. A recommendation is not a workflow. A concept is not a running system. Anyone buying consulting without defining an execution path buys paper.
How that is avoidable: Before any consulting engagement, clarify the question: who executes afterwards? With which budget? In which timeframe? If the answers are unclear, consulting is premature. First clarity on execution, then commission consulting - or directly engage an agency that does both.
The dual location as additional factor
An aspect often underestimated in model choice: spatial proximity. Self-evident with inhouse, not with agency and consulting. An agency in Villach with real presence in Graz has different options for Carinthian and Styrian clients than a German vendor with Swiss cloud.
What spatial proximity delivers: Strategy conversations on-site instead of just video, short-term workshop appointments, getting to know rooms and culture of the client, trust-building that is harder online.
What it does not deliver: Automatically better results. A poorly working local agency is still poor. But with comparable competence, the local vendor beats the remote vendor because relationship and fast availability count.
Anyone looking for a broader marketing plan that frames AI implementation as part of a larger system will find one in the 90-day plan book.
Hybrid models in practice
In reality the separation between the three models is rarely clean. Many companies combine - and do well by it. Three hybrid constellations that work particularly well in practice.
Hybrid 1: agency builds, inhouse runs
How it works: Specialized agency builds the workflows productively in 8 to 12 weeks. An inhouse person is onboarded in parallel and takes over running operations after 4 to 6 months. The agency stays available as escalation partner for complex adjustments.
When sensible: When the company already has an IT-affine person in marketing or sales who should build AI competence but does not plan full recruiting for an own role.
Advantage: Fast start at agency speed, long-term knowledge retention in the house, reduced running costs after handover.
Hybrid 2: consulting sharpens strategy, agency executes
How it works: A short IT consulting phase (4 to 8 weeks) develops strategic direction, tool recommendation and roadmap. Then consulting hands over to an agency for execution. Important: the consulting must know from the start that external execution will follow - otherwise concepts emerge that cannot be executed.
When sensible: When the strategic direction is genuinely unclear and several options need to be weighed. With clear use cases, the consulting phase is unnecessary.
Advantage: Strategic depth paired with operational execution power. Risk: handovers cost time and money - whoever does not coordinate both sides closely pays the friction.
Hybrid 3: inhouse core plus external specialists
How it works: An inhouse person is main responsible and works on the system daily. For special workflows (e.g. complex API integrations or new AI models) external specialists are added per project.
When sensible: Mid-market and larger companies with a clear AI growth plan over years.
Advantage: Best mix of knowledge retention and specialist access. Disadvantage: needs coordination competence in the house, which not every company brings.
What all three hybrids share
Clear responsibilities: Who decides when what? Who has the final word on tool choice? Who runs in daily operation? These questions must be clarified in writing before cooperation starts. Otherwise friction arises that eats up the hybrid advantages.
Handovers are mandatory checkpoints: Every hybrid model has changes between actors. These handovers are the most common weak points. Mandatory: documented handover protocols, training phases, clear escalation paths after the handover.
Trust is the foundation: Hybrid models often fail not on technology but on relationships between actors. Anyone not letting the people from consulting, agency and inhouse talk to each other has three silos instead of one system. Regular joint meetings, clear communication paths and mutual respect are not soft factors - they are the invisible substance every working hybrid model rests on.
Frequently asked questions on model choice
When does a specialized AI agency pay off for SMBs?
A specialized agency pays off for clearly bounded use cases (marketing automation, lead pipeline, reporting), with limited internal IT know-how, and when fast results matter more than maximum control. Strengths: experience from many similar projects, fast start, predictable running costs. Weakness: vendor dependency, knowledge sits external.
When does building an inhouse AI role pay off?
Inhouse pays off when AI applications become core business, the company has at least 100 employees and wants to invest long-term in own data competency. Advantages: deep process understanding, knowledge stays in the house. Disadvantages: recruiting takes 6 to 12 months, personnel costs 80,000 to 120,000 euros per year including ancillary costs, churn risk.
What sets an AI agency apart from IT consulting?
A specialized AI agency builds productive systems. Classical IT consulting creates concepts and recommendations. In the AI marketing context, the agency delivers running workflows; consulting typically delivers a roadmap. Consulting costs significantly more per day but is independent on tool recommendations.
How big is the risk with an AI agency?
The main risk is dependency. If the agency cancels the contract or works poorly, the workflow stalls. Mitigation: prefer daily-cancelable contracts, demand workflow documentation, regularly secure source code or workflow exports. Whoever agrees on this has reduced the dependency risk.
Which constellations fit which model?
SMBs with 10 to 50 employees and marketing needs: specialized agency. Mid-market 50 to 200 with clear AI strategy needs: mix of agency and external consulting. Mid-market above 200 with long-term data strategy: inhouse build-up plus selective external support. IT consulting pays off mainly in the concept phase, not in operations.
How do you choose an AI agency for marketing automation?
Six selection criteria: 1. demonstrable practical projects with references, 2. transparent communication on successes and failures, 3. tool stack fits your infrastructure, 4. daily-cancelable or short terms, 5. maintenance model in writing, 6. cultural fit in the first conversation.
Bottom line: the decision comes before the research
The most common order is reversed: company researches tools, compares vendors, gets three quotes - and only then notices that strategy is unclear. We recommend the reverse: first clarify what you actually want, what size you are and how your maturity situation looks. Then the model fits itself.
If you are an SMB with 30 employees and want a productive system fast, a specialized agency is the best fit. If you are mid-market with 200 employees and see AI as long-term strategy, build inhouse. If you are in the early strategy phase without clarity yet, you can start with brief consulting - assuming execution is already planned.
