Independent AI consulting & implementation, now taking on new projectsStart a conversation

AI agents that do real work
inside your business

Advice on where AI is worth using, then the agents and workflows to do it, built around the tools and data you already have.

Independent practice. The person who advises you is the person who builds it.

Example workflow: supplier invoice intake Illustration · not a client system
Email receivedTrigger

Starts when a supplier invoice arrives in the accounts inbox.

Frombilling@supplier.example
SubjectInvoice INV-2041
INV-2041.pdf
Waiting0.2 s
Invoice agentAI step

Reads the invoice and decides what needs checking.

ModelChosen per task
Tools
read_pdffind_pocheck_vendor
  1. Extract invoice fields
  2. Look up purchase order
  3. Verify supplier details
Waiting1.9 s
ERP lookupLookup

Finds the purchase order and compares amounts.

Order
PO-7731
Lines
3 of 3 match
Amount
agrees
Waiting0.4 s
CRM lookupLookup

Confirms the supplier and bank details on file.

Bank details
unchanged
Supplier
Example Supplier Ltd
Waiting0.3 s
Draft entryAction

Prepares the bill for your accounting system.

Amount
4,280.00
Account
6100
Due
in 30 days
Waiting0.6 s
Sign-offApproval

Nothing is posted until someone in Finance says yes.

ApproveEdit
Waiting—
PostedResult

Bill created, every step recorded for audit.

bill.createdaudit.logged
Waiting0.3 s
An example of the kind of system that gets built: an AI agent that does the reading and checking, connected to your existing tools, with a person approving before anything is posted.
What gets built

Systems your team uses every week, not demos

Each engagement ends with something running in your business. That usually means one or more of these building blocks, designed to fit how your team already works.

AI agents

Software that reads, decides on a next step and uses your tools, within limits you set.

Knowledge assistants

Answers from your own documents, with sources people can check.

Integrations

Connected to the tools you already use, with only the access each task needs.

Automated workflows

Fixed rules where rules work, and AI only for the steps that need judgment.

Testing on your real cases

A test set built from your own examples, run before launch and after every change.

Oversight built in

Approvals where they matter, and a log of what happened and why.

Interfaces shown are illustrations of what gets built, not screenshots of client systems.

Use cases

Workflows for every team

The best candidates are frequent tasks that involve reading, sorting, checking or drafting, where a person can catch a mistake before it matters.

Invoice intake that checks its own work

Connects to
Accounting software or ERP, accounts inbox, document storage
People decide
Every payment, posting and exception
More finance use cases
Agent description

Supplier invoices are read in any layout, matched to purchase orders and supplier records, and drafted as bills for a person to approve.

AI workflow
  1. Invoice email arrivesTrigger
  2. Extract fields from PDFAI step
  3. Match purchase orderLookup
  4. Verify supplier recordLookup
  5. Finance approvesApproval
  6. Bill posted, loggedResult
Draft billAwaiting approval
Supplier
Example Supplier Ltd
Invoice
INV-2041
Amount
4,280.00
Purchase order
PO-7731 · matched
Account
6100 · suggested
Flags
None

Example output, illustrative only

Requests that arrive complete and in the right place

Connects to
Email, forms, ticketing or project tools, ERP, spreadsheets
People decide
Changes to orders, schedules and commitments
More operations use cases
Agent description

Emails and forms become structured requests. Missing details are chased automatically, and the request is created in the right system for the right team.

AI workflow
  1. Email or form receivedTrigger
  2. Classify and extract detailsAI step
  3. Check required fieldsRule
  4. Ask for anything missingAction
  5. Request created and assignedResult
New service requestCreated · assigned
Requested by
Site manager (from email)
Category
Equipment repair
Location
Unit 4
Priority
High · suggested
Missing
Asset number, requested by reply
Assigned to
Maintenance queue

Example output, illustrative only

Tickets triaged, researched and drafted before anyone opens them

Connects to
Helpdesk, email, order or billing system, help center articles
People decide
Every reply sent, plus refunds and exceptions
More customer support use cases
Agent description

Each ticket is classified, the customer's order and history are looked up, and a reply grounded in your policies is drafted for an agent to check and send.

AI workflow
  1. Ticket createdTrigger
  2. Classify topic and urgencyAI step
  3. Look up order and historyLookup
  4. Draft reply from policyAI step
  5. Agent reviews and sendsApproval
Draft reply · ticket 1182Waiting for agent review

Hi Sam, thanks for getting in touch. Your order left our warehouse on Tuesday, and the carrier's tracking currently shows delivery on Thursday. If it hasn't arrived by Friday, reply here and we'll arrange a replacement.

[1] Order record[2] Delivery policy §2

Example output, illustrative only

Call notes that turn into CRM updates on their own

Connects to
CRM, calendar, email, call notes or transcripts
People decide
Anything sent to a prospect, plus pricing and commitments
More sales use cases
Agent description

After a call, notes or a transcript are summarized into structured CRM fields, next steps and follow-up tasks, ready for the salesperson to confirm.

AI workflow
  1. Call endsTrigger
  2. Summarize notesAI step
  3. Find account in CRMLookup
  4. Propose field updatesAI step
  5. Salesperson confirmsApproval
  6. CRM updated, tasks createdResult
Proposed CRM updateWaiting for confirmation
Account
Example Co.
Stage
Discovery to Evaluation
Next step
Send proposal by Friday
New contacts
Head of Operations
Follow-up
Task created for Thursday
Summary
Interested in invoice automation

Example output, illustrative only

Policy questions answered from your own handbook

Connects to
HR system, handbook and policy documents, chat tool
People decide
Any decision about a candidate or employee
More hr use cases
Agent description

An internal assistant answers questions about leave, benefits and procedures from your documents. It links to the exact section it used, and hands off to HR when it isn't sure.

AI workflow
  1. Question in chatTrigger
  2. Search policy documentsLookup
  3. Answer with citationsAI step
  4. Unsure? Route to HRRule
  5. Answer with source linksResult
Internal assistantAnswered with sources

Can I carry unused leave into next year?

Yes, up to five days, if your manager approves it before the end of the leave year. Anything above that is lost unless HR agrees an exception.

[1] Employee handbook §4.3[2] Leave request form

Example output, illustrative only

Integrations

Connects to the tools you already use

You don't need to replace your software. Agents work inside the systems you already pay for, through their official APIs wherever they exist.

Email & calendar
  • Microsoft 365
  • Outlook
  • Gmail
  • Google Calendar
CRM & sales
  • Salesforce
  • HubSpot
  • Pipedrive
  • Dynamics 365
Support
  • Zendesk
  • Freshdesk
  • Intercom
  • Help Scout
Finance & ERP
  • Xero
  • QuickBooks
  • NetSuite
  • Sage
  • SAP
Documents & knowledge
  • SharePoint
  • Google Drive
  • Dropbox
  • Notion
  • Confluence
Chat & work
  • Slack
  • Microsoft Teams
  • Asana
  • Jira
  • Monday.com
Data
  • Excel
  • Google Sheets
  • PostgreSQL
  • SQL Server
  • BigQuery
  • Snowflake
Automation platforms
  • n8n
  • Make
  • Zapier
  • Power Automate
AI models
  • OpenAI
  • Anthropic
  • Google Gemini
  • Azure OpenAI
  • Amazon Bedrock
  • Open-weight models

Examples of common tools with documented APIs. Product names are trademarks of their owners and are listed for identification only. Listing them does not imply any partnership or endorsement. If a system has an API, a database or reliable exports, it can usually be connected.

How it works

From first conversation to a system you own

Six stages, each with a clear output and a point where you decide whether to continue. You never commit to more than the next step.

  1. 01

    Assess

    Map where time goes and which tasks suit AI.

    Ranked shortlist
  2. 02

    Design

    Agree the workflow, controls and success measures.

    Written design & quote
  3. 03

    Build

    Build in short increments, connected to your tools.

    Working system in test
  4. 04

    Test on your real cases

    Run your own examples and fix what fails.

    Test results
  5. 05

    Deploy

    Roll out carefully, often with every output reviewed at first.

    Live system
  6. 06

    Hand over

    Train your team and hand over everything they need.

    Docs, runbook, access list

How each stage works

Security & data

Your data stays in your hands

AI systems touch sensitive information, so the controls are designed in from the start. Systems run in your accounts, with only the access each task needs, and people approve what matters.

Georgy Consulting is a one-person practice and holds no certifications such as SOC 2 or ISO 27001. Instead of badges, here is exactly how the work is done.

Your environment · your accounts
Your systemsCRM · ERP · drives · email
Scoped credentialsseparate, revocable
Agent or workflowruns in your account
Approvalsfor consequential actions
Audit loginputs, outputs, actions
Test setyour real cases
AI model providerbusiness or API terms, chosen and approved by you
Diagram: the agent, your systems, credentials, approvals, audit log and test set all sit inside your own environment and accounts. Only the data a task needs is sent to an AI model provider that you have approved.

Your accounts

Built and run in accounts your business owns, so billing, data and control stay with you.

Ownership001

Least access

Separate, revocable credentials with only the permissions a task needs.

Access002

Human approval

Sending, paying, deleting or changing key records can wait for a person.

Oversight003

Full audit trail

Inputs, outputs and actions recorded, so you can see what happened and why.

Audit004
Who does each stage
  • First conversationsame person
  • Assessment & advicesame person
  • Design & buildsame person
  • Testing & handoversame person
No account managers · specialists only with your written approval
Who you work with

One person, from first call to handover

Georgy Consulting is a one-person practice. The person who assesses your processes is the person who designs, builds and hands over the system, so nothing gets lost between advice and delivery.

  • No account managers or hand-offs to junior staff
  • Straight answers about what AI can't do, including when the answer is “not AI”
  • Recommendations based on the tools you already have
  • Everything documented and built in your accounts, so you're never dependent on one person
FAQ

Questions businesses ask first

More answers on the FAQ page.

We're not sure where AI would help. Is it too early to talk?

No, that is exactly what the consulting side is for. Bring a list of the tasks that take your team the most time, and the first conversation will narrow down whether any of them are good candidates.

Do we need to replace our existing software?

Usually not. Systems are designed around the tools you already use, connecting through their APIs or built-in automation features. If a tool can't be connected safely, you'll be told before any building starts.

Will an AI agent make mistakes?

Sometimes, yes. AI models can misread inputs or produce plausible but wrong output. That's why systems are tested on your real examples before launch, consequential actions can require a person's approval, and every step is logged so problems can be spotted and fixed.

Will our data be used to train AI models?

Major AI providers' business and API offerings generally say customer data isn't used for training by default, but terms vary and change. The chosen provider's current terms are reviewed with you before any of your data is processed.

How much does it cost?

It depends on scope. Every engagement is scoped and quoted in writing after the first call. The first call is free and takes 20 minutes.

Next step

Have a process you suspect AI could help with?

Start with a conversation about how the work is done today. You'll leave with an honest view of whether AI is worth pursuing for it.