About us
Retailbox builds automations that actually survive.
Since 2017, we've delivered AI automation, CRM, and workflow projects for B2B clients. We don't chase trends. We ship systems that keep running long after the project ends.
Meet the founder
Mikhail Oskola
Founder & AI Automation Architect
Nine years into building automations, I've learned one thing: the automation that works on demo day is not the one that still runs six months later. Retailbox exists because businesses deserve systems that survive API changes, team turnover, and real-world chaos — not just flashy demos.
I personally work on every project, from scoping to go-live. If a Retailbox automation fails at 3 AM, it's my phone that rings — and my standards that keep them from failing.
Mikhail on podcasts and stages
Mikhail talks about what happens to AI automation after the demo. He covers which projects are still running six months later, which ones break without anyone noticing, and how a small business can tell the difference before paying for one.
Short bio
Mikhail Oskola is the founder of Retailbox, an AI automation and CRM consultancy in Orlando, Florida. He started it in 2017 and still works on every project himself. On Upwork since 2019 he has completed 477 jobs with a 100% Job Success score and Top Rated Plus status.
Longer bio
Mikhail Oskola founded Retailbox in 2017. Since then he has built automations for businesses that don't have an engineer on staff: CRMs on Monday.com, HubSpot and Dynamics 365, and AI agents wired into the tools a team already uses.
Most of that work has gone through Upwork, where he has been since 2019. The profile shows 477 jobs, more than 17,900 hours and over $1M earned, with 100% Job Success and Top Rated Plus status. In the last 18 months Retailbox has worked with 68 clients.
His test for a project is whether it still runs six months after launch. Mikhail speaks English and Russian.
What he talks about
1.The automation that still works six months later
Demo day is easy. The hard part is month six, when an API changes, the person who asked for the automation leaves and nobody is watching the logs. Mikhail explains what he checks before taking a project, why he builds the predictable parts without AI where he can, and what monitoring a small business actually needs.
Good for: owners and ops leads deciding on their first AI project.
2.AI on top of the records you already have
A worked example from an appliance and HVAC repair company. Retailbox built a system that predicts which parts a technician should bring to a job. It was trained on 18,513 historical service records and built with Claude, n8n and ServiceTitan, plus dashboards on technician performance. The talk covers how old service history turns into a parts prediction and how it fits the tools the techs already use.
Good for: field-service, trades and home-services shows.
3.Clean the CRM before you automate it
For a US company, Retailbox moved 9,000 contacts and 1,000 accounts (40,000 rows) into Dynamics 365, cleaned them on the way, then set up sales pipelines that move deals between stages automatically. The talk covers what to fix in the data first and how to switch CRMs without stopping sales.
Good for: sales ops, RevOps and small-business shows.
Where he has been a guest
- AI Practic (YouTube, host Dmitry Gonchar), January 2026, in Russian. The episode is about why AI rollouts in business fail. Watch the episode
Booking
Mikhail records remotely from Orlando, Florida, on whatever platform your show uses.
Headshot: download (JPG, 1200×1200)
Invite Mikhail to your show
Send the show name and a link to a recent episode. We'll come back with dates.
What we do
- AI agents & LLM orchestration (OpenAI, Claude, LangChain, CrewAI)
- CRM implementation (Monday.com, HubSpot, Zoho, Dynamics 365)
- Workflow automation (n8n, Make.com, Zapier, Power Automate) — see our business process automation work
- Voice AI (Vapi, Retell, ElevenLabs)
- Full-stack development (Next.js, React, Node.js, Python)
- Data pipelines & integrations (REST, GraphQL, webhooks, MCP) — our AI integration services
Two of those lines have a page of their own. We wire OpenAI and Claude into systems a team already runs — see our AI integration services — and we automate the repetitive processes underneath them, which is business process automation: mapping a process, costing it, then removing the manual steps.
Our journey
- 2017
Retailbox founded
Started as a small CRM consulting practice.
- 2019
Expanded into workflow automation
Added Make.com, Zapier, and n8n. First multi-country clients.
- 2022
AI-first pivot
Started integrating GPT-4 and Claude into client workflows.
- 2024
CRM migrations at scale
For a US company: 9,000 contacts and 1,000 accounts (40,000 rows) migrated and cleaned into Dynamics 365.
- 2026
Managed service launch
Introduced Managed Automation plans — monitoring, free bug fixes, no vendor lock-in.
Recent videos
Walkthroughs and demos from @LetsAutomateThat — AI agents, voice AI, and automation case studies.
How to create an AI sales bot with ChatGPT-4 in just 7 minutes with NO CODE
Dec 2023 · 472 views
TARK Voice — Human-like AI solution for your business
Jul 2024 · 214 views
How to Increase Sales in Real Estate: Real AI Voice Integration
Jul 2024 · 97 views
How AI Voice Transforms Logistics: Trucking Company Communication
Jul 2024 · 76 views
Beyond work
Based in Orlando, Florida with my family. Being a dad changed how I think about what I build — every system has to actually work in real life, not just look good in a demo.
That's why “survives production” isn't just marketing copy at Retailbox. It's the bar.
Ready to talk?
30-minute free call. No obligation. We'll tell you honestly if we're the right fit.