# Generative AI Consulting That Ends With Working Software > Most AI consulting ends in a slide deck. What a good engagement covers, what it should cost, and the five questions that expose a deck seller before you pay. Source: https://llamapress.ai/generative-ai-consulting-that-ends-with-working-software | Updated: 2026-08-28 --- [LlamaPress Blog](https://llamapress.ai/blog) Most AI consulting ends in a slide deck. What a good engagement covers, what it should cost, and the five questions that expose a deck seller before you pay. [Back to all articles](https://llamapress.ai/blog) ![Darren David Spencer](https://llamapress-ai-image-uploads.s3.us-west-2.amazonaws.com/vdqignlvmc1rxv5n41gd4ihb12h5) Written by ### Darren David Spencer Ex-McKinsey Consultant | Operations Strategy Expert I made slide decks for a living. McKinsey trained me well, and the decks were good. The decks were also the product. Much artificial intelligence consulting still works that way today. The client pays a large fee, receives a roadmap, and owns no software at the end. If you are comparing AI consulting services right now, your main job is to avoid buying a document. This guide shows you what to buy instead. Good generative AI consulting covers four things: where AI fits your specific workflow, whether your data can support it, whether to build or buy each tool, and one pilot that ships during the engagement. Generative AI means software that writes text, code, or images on request. A pilot is a small first project with a measurable goal. The strategy phase should take two to four weeks. As of this writing, that phase runs from a few thousand dollars with an AI-native firm to six figures with a large consultancy. Judge every AI consulting company by one test. Ask what the previous client could use on the first day after the engagement ended. The gap between those two price points is usually the deck. ![Two consultants sketching an AI workflow plan on a whiteboard during a strategy session](https://images.unsplash.com/photo-1532622785990-d2c36a76f5a6?crop=entropy&cs=tinysrgb&fit=max&fm=jpg&ixid=M3w4NTU2MDF8MHwxfHNlYXJjaHwyfHxidXNpbmVzcyUyMHN0cmF0ZWd5JTIwY29uc3VsdGluZyUyMHdoaXRlYm9hcmQlMjBtZWV0aW5nfGVufDB8fHx8MTc4Nzg5NTcyNHww&ixlib=rb-4.1.0&q=80&w=1080) Photo by [Kaleidico](https://unsplash.com/@kaleidico?utm_source=leonardo_rails_app&utm_medium=referral) on [Unsplash](https://unsplash.com/photos/two-people-drawing-on-whiteboard-26MJGnCM0Wc?utm_source=leonardo_rails_app&utm_medium=referral) ## Why So Much AI Consulting Ends in a Deck The incentives explain the outcome. Most AI consulting firms bill by the hour or by the phase. Analysis hours are safe hours. A recommendation cannot crash in production, miss a deadline, or return a wrong answer to a customer. Working software can do all three. So the engagement model drifts toward what is safe to sell: interviews, market scans, maturity assessments, and a final readout. The firm leaves, and your team holds a PDF and a backlog. A deck is sometimes the right purchase. If you run a 2,000 person company, you may need an AI policy, a governance model, and board alignment before anyone builds anything. Regulated industries often need a risk review first. That work is real strategy work, and a large AI consultancy does it well. Buy the deck in those cases. If you run a 10 to 200 person company with a concrete process problem, you need something different. You need a working tool, and the strategy should be in service of shipping one. One more distinction saves buyers money. Some vendors that call themselves consultants are automation agencies that resell SaaS setups. I wrote a separate guide on [what an AI automation agency actually does](https://llamapress.ai/what-an-ai-automation-agency-actually-does-and-how-to-pick-one). Read that guide if the vendor's proposal lists subscription tools rather than analysis or code. ## What Generative AI Consulting Services Should Cover Strip the branding away and generative AI consulting services have four jobs. A good engagement does all four. A deck seller does the first one and stops. ### 1. Where AI fits your workflow The consultant should sit with your team and map the actual process: who enters what, where the handoffs happen, and where hours disappear. Good candidates for generative AI are narrow and repetitive. Drafting quotes from job notes, summarizing intake calls, and extracting data from emailed documents all qualify. A finding like "use AI in operations" is too vague to act on. A finding like "auto-draft the estimate email from the job record" is a spec. ### 2. Data readiness AI output is only as good as the records behind it. If your customer history lives in one operator's spreadsheet tabs, the model has nothing reliable to read. A good consultant checks where each needed fact lives today and what cleanup the pilot requires. This step often reveals the real project. Many companies need a database before they need a model. Our comparison of [Excel files and database-backed apps](https://llamapress.ai/excel-to-app) shows what that move looks like. ### 3. Build or buy, per use case Some needs are met by an off-the-shelf subscription, and an honest advisor says so. A generic meeting summarizer is a purchase, so do not commission one. A tool that quotes jobs with your price rules is a build, because no vendor sells your rules. The consultant should give a build or buy call for each use case, with the monthly cost of each path. ### 4. One pilot that ships The engagement should end with one pilot live in your business, used by your team, with a number attached. Pick the metric before the build starts: hours saved per week, response time, error rate on quotes. One shipped pilot teaches you more about AI in your business than any market scan. The pilot also tests the vendor. A firm that cannot ship a small tool will not ship a large one. ## What AI Strategy Consulting Should Cost and Produce Prices vary widely, so treat these bands as orientation, not quotes. As of this writing, large firms price AI strategy consulting in the six figures for enterprise clients. Boutique AI consulting firms commonly land in the tens of thousands for a strategy phase. An independent AI business consultant often charges $100 to $300 per hour. AI-native companies like ours fold the strategy work into a build priced in the low thousands, because the analysis and the software come from the same effort. Whatever you pay, the strategy phase should produce four artifacts. First, a workflow map that names the specific steps where AI applies. Second, a data readiness assessment with the cleanup work listed. Third, a build or buy decision for each use case with running costs. Fourth, a scoped pilot with an owner, a deadline measured in weeks, and a success metric. If a proposal promises a "transformation roadmap" and none of those four, you are reading a deck outline. Company size changes the math. Under roughly 50 employees, a separate strategy engagement rarely pays for itself. The analysis fits in a week, and the budget belongs in the pilot. I wrote a guide to [the practical path to automation for small businesses](https://llamapress.ai/ai-consulting-for-small-businesses-the-practical-path-to-automation) that covers that case in depth. This page assumes you have a process, a team, and a real decision to make between vendors. ## Five Questions That Expose a Deck Seller Every AI consulting company sounds identical in the pitch. The same case studies, the same maturity curve, the same urgency. These five questions separate the builders from the presenters. 1. **Who on the team writes code?** Ask for the names and roles of the people assigned to you. An engagement staffed only with strategists and analysts will end in analysis. 2. **Can the pilot go live inside the engagement?** The only good answer is yes, with a date. "Phase two" means the shipping risk stays with you. 3. **Who does the AI integration consulting after the strategy phase?** Integration means connecting the AI tool to your live systems and data. If the answer is "an implementation partner," you are buying a handoff, and handoffs are where projects die. 4. **What did your last client use on day one after you left?** Listen for a named tool in daily use. Listen with suspicion to a list of documents. 5. **Who owns what you produce?** Code in your repository and data in your database, in writing. A consultant whose deliverables live in their systems has made leaving expensive. None of these questions require technical knowledge. They only require asking, before money moves, what exists when the engagement ends. ## How We Consult: We Ship the Pilot LlamaPress is an AI consulting company of the fourth-artifact kind. We skip the deck and start at the pilot. Our AI coding agent, Leonardo, builds a working database-backed web app from the process you already run. Your current spreadsheet is usually the best spec, because your team already encoded the workflow in it. The strategy questions still get answered: workflow fit, data readiness, build or buy. They get answered in working software, in days, priced in the low thousands of dollars. > "Instead of my team having to do all that, why can't it just all be automated?" That question came up on a real call. The answer became a build list, not a deck. Our answers to the five questions above, on the record: the agent and our engineers write code from day one. The pilot goes live inside the engagement. We do the integration work ourselves. The deliverable is an app your team signs into. The code lives in your own GitHub account, and your data sits in a database you can export. And to be plain about fit: we do not write enterprise AI policy, run change management programs, or manage anyone's ad campaigns. If your need is a 200 slide alignment exercise, hire a strategy firm. If your need is software that works, the [services page](https://llamapress.ai/services) shows how an engagement runs. Strategy That Ships ### Skip the deck. Start at the pilot. Bring the workflow you want AI to carry, or just bring the spreadsheet that runs it. We will show you a working pilot before you commit to anything. [See how we consult](https://llamapress.ai/services) [Describe your workflow](https://llamapress.ai/) Prefer to start with a message? [Contact us](https://llamapress.ai/contact) and tell us what you run today. ## Frequently Asked Questions ### What do AI consulting services include? A complete engagement includes four things: a map of where AI fits your workflow, a data readiness assessment, a build or buy decision for each use case, and one pilot that ships during the engagement. Many firms deliver only the analysis. Confirm before signing that a working pilot is inside the scope. ### How much does generative AI consulting cost? As of this writing, large consultancies price enterprise engagements in the six figures. Boutique firms commonly land in the tens of thousands for a strategy phase. Independent consultants charge $100 to $300 per hour. AI-native builders fold strategy into a pilot priced in the low thousands of dollars. ### What is the difference between AI strategy consulting and AI integration consulting? AI strategy consulting decides what to do: which use cases, which data, build or buy. AI integration consulting makes a chosen tool work inside your live systems and data. Deck sellers stop at strategy and hand integration to someone else. Prefer a vendor that does both, so one party owns the outcome. ### Should I hire an AI consulting firm or an independent AI business consultant? Pick by what must exist when the engagement ends. A solo consultant suits advice, vendor selection, and team training. A firm suits work that needs designers, engineers, and delivery on a deadline. In both cases, apply the same test: ask who writes the code and whether the pilot goes live inside the engagement. ### How do I choose an AI consulting company? Ask five questions: who on the team writes code, can the pilot go live inside the engagement, who handles integration, what the last client used on day one after exit, and who owns the code and data. Then weigh price against shipped output. A working pilot beats any roadmap.