All Case Studies
HOW WE WORK
How Dcrayon plans an AI project around one business number
A short film on how we work with clients: who you talk to, how we report, and why every project starts with a free review before any build.
Three tools behind every AI project
The review that finds the right first use case, the plan that ties it to one number, and the in-house AI tools we test with.
Growth Intelligence (AI review)
A free 150-point review of your processes, data and software. It ranks where AI can save hours or add sales first, and where it should wait.
Dcrayon Growth Engine (AI plan)
A written 90-day plan with one target, such as hours saved or enquiries answered, plus the points where a person checks the AI.
Dcrayon AI (in-house toolkit)
The agents, content checks and publishing automation we run in our own agency every day, adapted for each client build.
How Dcrayon delivers an AI project
Four steps. Each one ends with something you can see and test.
Step 1: Review
We map one workflow with your team, count the hours and errors it costs today, and collect 30 to 50 real examples: emails, chats, documents or orders.
Step 2: Prototype
We build the smallest working version on your examples and compare models from OpenAI, Anthropic and Google on accuracy and cost per task.
Step 3: Launch
The AI goes live with limits: what it may do alone, what needs a person's approval, and a log of every action. Your team gets a short handover session.
Step 4: Improve
Each week we review the cases it got wrong, update instructions and tests, and report the one number from your plan in plain words.
Get my free AI review in one business day
A free 150-point review of where AI fits your business, plus the 90-day plan to start.
+187%
Median sales lift, first year
Our Artificial Intelligence services
Six AI programmes, from a first consulting session to agents that act inside your software. Most clients start with one and add the next once it pays back.
AI work you can check for yourself
A live app and a live store, both with public case studies.
Why choose Dcrayon for AI work
We use it before we sell it
Our CMS takes drafts from ChatGPT through a connector we built, checks every save for banned words, and posts to five social channels on publish.
A named person on every AI step
AI can draft, sort and suggest. Anything that touches your brand, legal terms or money waits for a person, and every action is logged.
Tested on your examples first
Before launch we score the AI on real cases from your business, so you see accuracy and cost per task before you commit.
Fits the software you already run
We connect AI to your CMS, CRM, store or help desk through their APIs. Python, Laravel, Node.js and Flutter are our everyday stack.
We have built software and marketing systems for brands since 2016, and we run AI in our own agency every day before we build it for anyone else.
Your AI review costs nothing, and you keep the 90-day plan either way.
We have worked with 485+ brands. The review shows which workflow to automate first, what the AI should cost to run each month, and what a person still needs to check.
A free 150-point review of your workflows, data and software
One first use case, ranked by hours saved or sales added
A model and running-cost estimate before any build starts
Month to month after the first 90 days, no yearly contract
Read before you start an AI project
Three guides from our team on AI agents, model costs, and how AI tools decide which brands to name.
Enterprise AI Agent Architecture: Tool-Use, MCP, and the Production Discipline
Reference architecture for production AI agents in 2026. Tool design, Model Context Protocol, sandbox isolation, eval suite, and the cost + safety discipline.
Read More about Enterprise AI Agent Architecture: Tool-Use, MCP, and the Production Discipline
How to Cut Your LLM Costs by 50% in 2026: Caching, Batching, and Model Tiering
Practical guide to cutting LLM API costs by 50%+ in 2026. Prompt caching, batch APIs, model tiering, response truncation, and the monitoring discipline.
Read More about How to Cut Your LLM Costs by 50% in 2026: Caching, Batching, and Model Tiering
How to get cited by ChatGPT, Gemini, and Perplexity in 2026
Answer engines pick a handful of sources per query and ignore the rest. Here is how brands earn citations across ChatGPT, Gemini, Perplexity, and Google AI Overviews. with examples and a 14-day audit you can run yourself.
Read More about How to get cited by ChatGPT, Gemini, and Perplexity in 2026
Question & Answer
Frequently asked questions
Plain answers to what business owners ask us before an AI project.
Dcrayon offers six AI services: AI consulting, AI development, AI agent development, AI chatbot development, generative AI integration and AI automation. We also run AI search optimisation, which helps your brand get named in ChatGPT and Google AI answers. Most clients start with a free review that picks one first use case.
Each AI project is priced as a fixed quote, written up once the free review has scoped the first workflow and agreed before any build begins. The quote has two parts: the build, and the monthly running cost of the AI models. We estimate that running cost per task on your own examples, so the monthly bill holds no surprises.
We work with models from OpenAI, Anthropic (Claude) and Google (Gemini), and with open models when data must stay on your own servers. We pick per task after testing a few on your real examples, and we design the build so you can switch provider later without starting again.
Yes, when the setup is right. We send a model only the data a task needs, use business API terms under which your data is not used for training, keep keys and tokens encrypted, and log every action the AI takes. For sensitive data we can run an open model on your own infrastructure.
You see a working version on your own examples before anything goes live, and your 90-day plan names the date for that demo. Hours saved show up as soon as the first workflow is live. Larger gains come from the second and third workflows, once your team trusts the first one.
No. In our projects AI takes the repeat steps, such as sorting enquiries, drafting replies or checking listings, and your people keep the decisions. We set clear approval points, so a named person checks anything that touches your brand, legal terms or money.
No. The first 90-day cycle covers the review, the build and the launch of one workflow. After that you go month to month, and every agreement lets you leave. You keep the code, instructions and test sets we build for you.
Get my free AI review
A 150-point review of where AI fits, plus the 90-day plan to start.
Start with one workflow, not a big AI programme
Tell us the task that takes your team the most hours each week. We will show you, on your own examples, what AI can and cannot do with it.
If AI is not the right answer yet, we will say so and tell you what to fix first.