---
type: "BlogPosting"
title: "AI Strategy in 90 Days: What AI Consulting Actually Delivers"
description: "Use-case list, data check, prototype, 12-month roadmap, EU AI Act classification and business case: what should exist after around 90 days of AI strategy consulting, with phase plan, roles and quality criteria."
resource: "https://www.contextstudios.ai/blog/ai-strategy-90-days"
language: "en"
tags: ["AI Strategy", "AI Consulting", "AI Roadmap", "EU AI Act", "SMEs"]
generated:
  by: "process:contextstudios-md/1"
  at: "2026-11-13T14:43:27.839Z"
status: "stable"
---

# AI Strategy in 90 Days: What AI Consulting Actually Delivers

Published: 2026-11-11
Tags: AI Strategy, AI Consulting, AI Roadmap, EU AI Act, SMEs

![AI Strategy in 90 Days: What AI Consulting Actually Delivers](https://wary-platypus-754.convex.cloud/api/storage/8b79dd55-94e6-43a3-9bed-8e79ff23abdc)

After roughly 90 days of AI strategy work, six results should be on the table: a prioritised list of use cases, a data and systems check, one validated prototype or a reasoned decision against a proof of concept, a 12-month roadmap, a classification under the EU AI Act and internal governance, and a business case for each prioritised use case. An AI consultancy that delivers only slides and recommendations after three months has skipped the most important part of the work: testing whether the ideas actually work with your data, your systems and your budget.

The 90 days are a typical frame, not a fixed commitment. How long it really takes depends on how quickly data becomes available, how available your contacts are and how many use cases are seriously examined. This article describes what is produced in each phase, who is needed on your side and how to tell solid [AI consulting](/ai-consulting) from a weak offer.

## What should be on the table after 90 days

An AI strategy is not a document but a series of decisions based on tested assumptions. The following six deliverables make those decisions traceable:

1. **Prioritised use-case list.** All collected ideas, rated by benefit, effort, data availability and risk. Typically three to five initiatives end up at the top; the rest are deliberately deferred, with reasons.
2. **Data and systems check.** Which data sits where, in what quality and with which access rights? Which systems (ERP, CRM, DMS, ticketing) need to be connected, and do they have interfaces for it?
3. **One validated prototype or a PoC decision.** For the most important use case, we test whether it holds up technically. The result is either a working prototype with measured results or a documented decision on why a proof of concept does not (yet) make sense.
4. **12-month roadmap.** Initiatives with sequence, dependencies, milestones, effort classes and responsibilities, as a basis for budget and planning.
5. **Governance and EU AI Act classification.** Each prioritised use case is assigned a risk class, data protection questions are settled and there are internal rules for using AI.
6. **Business case per use case.** Expected benefit as a target value with disclosed assumptions, plus costs for implementation and operation and the metric by which success will be measured.

If one of these elements is missing, the strategy is not wrong, but it is incomplete. Points 2 and 3 are the ones most often missing, because they require real work with your data.

## The phase plan: days 1–30, 31–60, 61–90

The following plan shows how the work is typically split across three phases. Each phase ends with deliverables you can sign off before the next one begins.

| Phase | Focus | Deliverables at the end of the phase |
|---|---|---|
| Days 1–30 | Discovery and assessment | Rated use-case list, data and systems check, initial risk classification, choice of pilot use case |
| Days 31–60 | Prototype or proof of concept | Working prototype with test data, defined success criteria, measured results, effort estimate for production |
| Days 61–90 | Decision and roadmap | Go/no-go decision, business cases, 12-month roadmap, governance rules, decision paper for management |

### Days 1–30: Discovery and assessment

The first phase starts with a workshop in which business units and IT jointly collect and roughly rate AI ideas. The assessment follows: we look at the data situation, system landscape, processes and skills. For a full assessment with a draft of the prioritised roadmap, we typically plan two to four weeks. At the end of this phase it is clear which use case will be tested first, and why that one.

### Days 31–60: Prototype or proof of concept

In the second phase, the most important use case is built, not just described. Our target is a usable prototype in about four weeks. Success criteria are set beforehand: what answer quality, what processing time, what error rate must the prototype reach for further development to be worthwhile? We work with anonymised or synthetic sample data until production data is released.

### Days 61–90: Decision and roadmap

The third phase turns results into decisions. The prototype is evaluated against the success criteria, the business cases are updated with the measured values and the roadmap orders all initiatives for the next 12 months. Governance rules and a decision paper for management are added. Afterwards you can implement internally, with a partner of your choice or with us.

## Who is needed on your side

An AI strategy does not come about without your team. These roles should be named before work begins:

- **Sponsor from senior management.** Makes the go/no-go decision and releases budget. Few meetings, but at the right points.
- **Person responsible for the initiative.** Coordinates meetings, data access and alignment, ideally with a few hours per week.
- **Subject-matter experts.** One or two people who know the process AI is meant to improve. Without them, every assessment stays theoretical.
- **Technical contact.** Knows systems, interfaces and security requirements and can grant access.
- **Data protection and, where applicable, the works council.** Should be involved early. If an AI system can evaluate employees' behaviour or performance, the co-determination right under [Section 87(1) no. 6 of the German Works Constitution Act (BetrVG)](https://www.gesetze-im-internet.de/betrvg/__87.html) applies.

On our side, you speak directly with founder Michael Kerkhoff, who is also technically accountable for the recommendations. Specialised freelancers join as needed.

## EU AI Act: what applies in November 2026

Classification under the EU AI Act belongs in every AI strategy, because it determines the documentation and testing effort a use case will require. The picture shifted in 2026: the Digital Omnibus on AI, [Regulation (EU) 2026/1744](https://eur-lex.europa.eu/eli/reg/2026/1744/oj), published in the Official Journal on 24 July 2026, changed several deadlines of the [AI Act, Regulation (EU) 2024/1689](https://eur-lex.europa.eu/eli/reg/2024/1689/oj). According to the [implementation timeline of the European Commission's AI Act Service Desk](https://ai-act-service-desk.ec.europa.eu/en/ai-act/timeline/timeline-implementation-eu-ai-act), the current position is:

| Date | What applies |
|---|---|
| 2 February 2025 | Prohibited practices and AI literacy provisions (Art. 4) |
| 2 August 2025 | Obligations for providers of general-purpose AI models |
| 2 August 2026 | Most of the rules, including transparency obligations under Art. 50; enforcement begins |
| 2 December 2026 | New prohibitions (including non-consensual sexual deepfakes); transition period under Art. 50(2) for certain systems placed on the market before 2 August 2026 |
| 2 December 2027 | High-risk obligations for Annex III systems (previously 2 August 2026) |
| 2 August 2028 | High-risk obligations for AI in Annex I products (previously 2 August 2027) |

On AI literacy: the Omnibus has softened Art. 4. Providers and deployers must support the development of AI literacy rather than ensure a specific level, as the [Future of Privacy Forum](https://fpf.org/blog/the-ai-act-implementation-timeline-what-changes-under-the-ai-omnibus) summarises. For the strategy, this means training remains sensible and should be on the roadmap, but proving a minimum level per person is no longer an obligation.

For most use cases in mid-sized companies, such as knowledge search in manuals, quote preparation or document classification, the classification is usually minimal or limited risk. For chatbots or AI-generated content, the transparency obligations under Art. 50 are added. It is different for Annex III applications, for example in HR for candidate selection or performance evaluation. The postponement to December 2027 is no reason to start such projects without governance: the obligations are coming, and documentation can only be produced retroactively with considerably more effort. This section is an orientation, not legal advice.

## Typical pitfalls

These are the patterns we see most often in AI strategy projects:

- **Too many use cases at once.** Testing twenty ideas in parallel means testing none of them thoroughly. Better: take one use case through to a decision, then the next.
- **Data is checked too late.** The best idea fails if the data is scattered, incomplete or not legally usable. The data check belongs in the first phase, not in implementation.
- **Success is not defined.** Without a metric set in advance, every prototype gets declared a success. Define what "good enough" means before building.
- **The business unit is missing.** Strategies developed only between management and IT often fail on acceptance in daily work.
- **Strategy without an implementation path.** A roadmap without responsibilities, effort classes and a budget frame remains a wish list.
- **Model choice as the starting point.** Whether Claude, GPT, Gemini or an open model such as Llama or Mistral fits follows from the use case, data protection and cost, not the other way round.

## Good or weak AI strategy consulting: how to tell the difference

| Criterion | Good strategy consulting | Weak strategy consulting |
|---|---|---|
| Result per phase | Named deliverables that you sign off | "Insights" and slides without clear acceptance |
| Feasibility | Tested with your data and systems | Assumptions from industry studies |
| Prototype | Built and measured, or rejected with reasons | Not planned or outsourced to third parties |
| Pricing model | Fixed price after scoping, scope clear upfront | Open hourly billing without a cap |
| Technology | Vendor-neutral, decision justified | Committed to one vendor from the start |
| Handover | Roadmap your team or another partner can implement | Dependence on the consultant for every next step |
| Regulation | EU AI Act classified per use case, with current deadlines | Generic reference to "compliance" |

A good test question in the first meeting: "What exactly will we hold in our hands after 30 days?" Anyone who has no concrete answer will not have one after 90 days either. How pure consulting differs from consulting with in-house implementation is shown in our [comparison End-to-End vs. AI Consulting](/comparisons/end-to-end-vs-ai-consulting). If you are weighing an agency against an individual, the article [AI agency vs. freelancer](/ai-agency-vs-freelancer) helps.

## Workshops as the entry point: fixed prices and scoping

Not every company needs 90 days straight away. With us, the entry point is always a fixed-price workshop:

- **Light Discovery** (½ day, remote): €1,500 net. Collect and rate ideas, first prioritisation.
- **Strategy Day** (1 day, remote or on site): €2,500 net. Rate and prioritise use cases with a feasibility and ROI matrix.
- **Prototyping Sprint** (2 days, on site recommended): €4,500 net. Additionally one or two clickable prototypes and a go/no-go decision.

Part of the workshop result is a 30/60/90-day plan with responsibilities. This plan is the starting point for the phases above: it shows which parts your team can take on itself and where support makes sense. For anything beyond the workshop: fixed price after scoping, quote within 48 hours. We split larger projects into milestones with their own fixed prices. That way you know before each phase what to expect. What AI projects typically cost after the strategy is explained under [AI development cost](/ai-development-cost). Details on the formats can be found under [AI workshops](/services/ai-workshop).

If you want to assess in advance where your company stands, you can use the [AI maturity self-test in the SME guide](/guides/ki-mittelstand-2026#self-test): seven questions, with pointers to sensible next steps. The same guide covers formats and funding options for mid-sized companies, and on site in Berlin-Charlottenburg we work under [AI Consulting Berlin](/ai-consulting-berlin).

## Frequently asked questions

### How long does it take to develop an AI strategy?

A workshop takes half a day to two days. An assessment with a prioritised roadmap typically needs two to four weeks. If a prototype is also built and evaluated, around 90 days is a realistic frame. What matters most is the availability of your contacts and access to data.

### What does AI strategy consulting cost?

The entry point is a fixed-price workshop: €1,500, €2,500 or €4,500 net, depending on the format. For assessment, prototype and roadmap: fixed price after scoping, quote within 48 hours. We do not offer open hourly billing.

### Do we need an AI roadmap if we already have a concrete use case?

Not necessarily. If use case, data and success criterion are clear, a proof of concept is often the more direct route. A roadmap pays off as soon as several initiatives compete for budget and staff or management expects an overall plan.

### What happens if the prototype does not meet the success criteria?

Then that is a result, not a failure. With limited effort you have prevented budget from flowing into an initiative that would fail on data, quality or cost. The roadmap is adjusted accordingly, and the next use case moves up.

### Do we have to commission Context Studios for implementation?

No. The roadmap is structured so that your own team or another partner can work with it. With full payment you receive the exclusive rights of use to code and prototypes from workshops. If you wish, we also implement the initiatives ourselves. You can arrange a free 30-minute initial call via the [contact page](/contact).

## Sources

- Regulation (EU) 2026/1744 (Digital Omnibus on AI), Official Journal of the EU, 24.07.2026: https://eur-lex.europa.eu/eli/reg/2026/1744/oj
- Regulation (EU) 2024/1689 (AI Act): https://eur-lex.europa.eu/eli/reg/2024/1689/oj
- European Commission, AI Act Service Desk, implementation timeline (accessed 02.11.2026): https://ai-act-service-desk.ec.europa.eu/en/ai-act/timeline/timeline-implementation-eu-ai-act
- Future of Privacy Forum, "The AI Act Implementation Timeline: What Changes Under the AI Omnibus?": https://fpf.org/blog/the-ai-act-implementation-timeline-what-changes-under-the-ai-omnibus
- Section 87 German Works Constitution Act (BetrVG): https://www.gesetze-im-internet.de/betrvg/__87.html


## Related

- [AI Consulting](https://www.contextstudios.ai/ai-consulting.md)
- [AI Agency](https://www.contextstudios.ai/ai-development-company.md)
- [AI Development](https://www.contextstudios.ai/ai-development.md)
- [AI Services](https://www.contextstudios.ai/ai-development-services.md)