Bubbles
dotpy.tech Delivers a 4-Day Applied AI Program to Bubbles
From conversational fluency with AI to live operational agents — equipping one of Egypt’s leading baby-care brands to run on AI, not just talk about it.
Over four intensive days, dotpy.tech partnered with Bubbles — one of Egypt’s most recognised baby and child care brands — to move its teams from curiosity about artificial intelligence to genuine operational capability. The program was deliberately built around the company’s own workflows rather than generic demonstrations, so that every prompt, agent, and automation created in the room mapped directly onto work the Bubbles team does every week.
Selected moments from the four-day dotpy.tech × Bubbles corporate AI program.
Module One · Day 1
Communication with AI
The program opened where every successful AI transformation actually begins: with language. Before a single automation was built, dotpy.tech’s trainers addressed the single largest source of wasted AI effort inside companies — vague, under-specified instructions. Participants learned that a large language model is not a search engine to be queried but a capable collaborator to be briefed, and that the quality of any AI output is bounded almost entirely by the quality of the brief it receives.
The module worked through the anatomy of a professional prompt: establishing role and expertise, supplying business context, defining the exact output format, setting constraints, and providing worked examples the model can pattern against. Teams practised on live Bubbles material — product descriptions, customer enquiries, internal reports, supplier correspondence — rewriting weak instructions into precise, reusable briefs and watching output quality shift in real time. Iteration was treated as a discipline rather than a failure: participants learned to diagnose why an answer missed, then correct the instruction instead of abandoning the tool.
Crucially, the session also covered judgement. Participants were trained to recognise where AI is genuinely reliable, where it requires verification, and where confidential or regulated information must never be pasted. By the close of Day 1, the Bubbles team shared a common vocabulary for working with AI and a personal library of prompts already producing usable output — the foundation everything that followed was built on.
Module Two · Day 2
Mini Operational Agents
Day 2 moved from conversation to delegation. Rather than starting with sprawling, enterprise-scale AI ambitions, dotpy.tech introduced the concept of the mini operational agent: a small, tightly scoped AI assistant built to perform one recurring business task exceptionally well, every single time, with behaviour fixed by deterministic instructions rather than left to improvisation.
Participants learned the architecture behind these assistants — a clear role definition, hard behavioural rules, a bounded knowledge base of company documents, and a strictly defined output structure. The distinction between a general-purpose chatbot and a purpose-built operational agent became immediately tangible: the former is a conversation, the latter is a dependable component of a workflow whose output can be trusted, reviewed, and forwarded without rework.
Working in teams, the Bubbles cohort identified and built agents against their own operational pain points — candidate and CV screening, customer enquiry classification and response drafting, product and marketing copy generation aligned to brand tone, internal report summarisation, and standard-operating-procedure lookups for new staff. Each team left with a functioning assistant and, more importantly, with a repeatable method for spotting the next task worth delegating.
- Customer enquiry triage — classifies incoming messages and drafts an on-brand first response.
- Product content assistant — generates descriptions and campaign copy in the Bubbles voice.
- Report summariser — converts long operational updates into executive-ready briefs.
- SOP & policy assistant — answers internal process questions from approved company documents only.
Module Three · Day 3
The Claude Ecosystem
Day 3 took the cohort deep into a single professional-grade platform. Rather than a shallow tour of a dozen tools, dotpy.tech dedicated a full day to the Claude ecosystem — the environment where the Bubbles team would actually do sustained work. Participants moved beyond the chat window into the capabilities that separate casual AI use from genuine productivity: Projects for persisting company context across conversations, custom instructions for locking in tone and standards, and file-based work for analysing real documents and spreadsheets.
The session covered Artifacts for producing shareable deliverables — dashboards, reports, one-pagers, and internal tools generated directly from a brief — and Skills for packaging a repeatable company procedure into something any team member can trigger reliably, without rebuilding the instruction each time. Teams also explored connectors, seeing how AI can work directly against the systems where their data already lives rather than requiring endless copy-and-paste.
Governance ran alongside capability throughout. Participants examined workspace structure, what belongs in shared versus personal context, how to keep sensitive commercial and customer data appropriately handled, and how to build institutional knowledge that survives staff turnover. The outcome was a team that understood not just which buttons to press, but how to organise an AI workspace that a growing company can actually sustain.
Module Four · Day 4
Automation
The final day connected everything into systems that run without supervision. dotpy.tech drew the critical line between AI assistance — a human asking, a model answering — and AI automation, where a trigger fires, work is completed, and a result is delivered with no one in the loop. For a consumer brand operating at Bubbles’ volume, that distinction is the difference between a useful tool and measurable operational leverage.
Participants learned to map a process before automating it: identifying the trigger, the decision points, the data sources, the actions, and the human checkpoints that must be preserved. They then built working scenarios combining no-code automation platforms with AI reasoning steps and business systems such as spreadsheets, email, forms, and messaging — covering rule-based flows, AI-enhanced flows, and fully agentic workflows in which the AI selects its own next action within defined boundaries.
The day closed on sustainability rather than novelty. Teams worked through error handling, notification and escalation design, logging, and cost awareness — the unglamorous engineering that determines whether an automation is still running in six months or quietly abandoned after two weeks. Each group left with live automations in place and a prioritised backlog of the next processes to tackle, scored by effort against hours saved.
- Rule-based flows — deterministic, high-volume tasks with no judgement required.
- AI-enhanced flows — classification, extraction, summarisation, and drafting inside a fixed pipeline.
- Agentic workflows — multi-step processes where the AI chooses its next action within set guardrails.
- Human-in-the-loop checkpoints — approval gates on anything customer-facing or commercially sensitive.
“We came in expecting a lecture about the future. We left on day four with agents and automations already running on our own work — that is a completely different kind of training.”
Training Participant — Bubbles
Key Outcomes
What the Bubbles team walked away with after four days.
- ✓A shared, practical standard for briefing AI effectively — with a personal prompt library per participant, built on real Bubbles material.
- ✓Functioning mini operational agents covering enquiry triage, content generation, reporting, and internal SOP lookups.
- ✓Working command of the Claude ecosystem — Projects, custom instructions, file analysis, Artifacts, and Skills — organised for team-wide reuse.
- ✓Live automations in production, spanning rule-based, AI-enhanced, and agentic workflow patterns.
- ✓A prioritised automation backlog, scored by implementation effort against projected hours saved each month.
- ✓Clear internal guardrails for data handling, verification, and human approval on anything customer-facing.
Partner with dotpy.tech for your next corporate AI program
dotpy.tech is an AI and Data Science training academy based in Cairo, serving organisations across Egypt and the wider MENA region. We design and deliver applied AI programs — prompt engineering, AI agents, no-code automation, and AI for business — built entirely around each client’s own workflows, data, and commercial objectives.
Every engagement is measured the same way: what the team can do differently on Monday morning that it could not do the Friday before.
Request a training proposal →