Esther Enuma / AI Automation & Digital Systems
status: open for AI automation & digital systems projects

I build AI systems that turn customer conversations into business outcomes.

I design and deploy voice agents, lead-response workflows, support systems, and conversion-focused web experiences for service businesses and growing teams. Every system is tested, documented, and built for the client to own.

01 / Work

Three working systems, tested around the failures that usually break automation.

SYSTEM 01
shippedvoice

AI Voice Receptionist

conversation → live availability → booking → owner summary

Before

After-hours and overflow calls reach voicemail. Staff must qualify every caller, check the calendar, create the appointment, and copy the details into a separate record.

After

A browser-based voice receptionist qualifies the caller, checks live availability, books the selected slot, logs the call, and sends the owner a written summary.

call answeredqualifycheck calendarbooklogowner summary

Key decision

If the calendar or another connected service is unavailable, the agent does not guess. It captures a callback request so a technical failure becomes a recoverable lead rather than a false booking.

1 call
from spoken enquiry to a confirmed appointment, with the call details recorded for the owner.
vapi · n8n · google calendar · google sheets · gmail
SYSTEM 02
shippedlead ops

Lead Response Engine

public form → storage → ai classification → reply → owner alert

Before

Enquiries are checked manually, answered hours later, and tracked inconsistently. High-intent leads are easy to miss when the owner is busy or offline.

After

Every web enquiry is stored, classified by intent and urgency, scored for purchase readiness, answered with a tailored first response, and escalated to the owner when intent is high.

capturestore firstclassifyvalidatereplyalertupdate status

Key decision

The lead is written to storage before the model runs, and model output is validated and safely defaulted. An AI failure may delay classification; it never loses the enquiry.

< 60 sec
to deliver a useful first response in end-to-end testing while preserving the lead and its status.
n8n · google sheets · gmail · ai classification · javascript · json
SYSTEM 03
shippedsupport agent

Customer-Support AI Agent

chat → memory → knowledge tool → grounded response

Before

Support teams repeatedly answer the same policy, delivery, refund, product, and opening-hours questions, while customers wait for someone to search the approved information.

After

An AI agent remembers the conversation, calls a dedicated Airtable knowledge tool, and returns one approved answer. Unsupported questions are not answered from guesswork.

chatmemoryknowledge toolairtable searchone resultgrounded reply

Key decision

Knowledge retrieval lives in a separate tool workflow that consolidates multiple Airtable records into one structured result. That prevents duplicate agent runs and keeps the main agent simple.

1 → 1
one customer question produces one approved response instead of several Airtable items triggering duplicate answers.
n8n · airtable · openai · ai agent · conversation memory · tool-calling
02 / Web experiences

Customer-facing websites designed to make the business clear, credible, and easy to act on.

I also design and develop the digital experiences through which customers discover a brand, understand its offer, and take the next step.

Lagos Luxe Interiors website homepage
WEB 01
liveinteriors

Lagos Luxe Interiors

A refined lead-generation website for an interiors and custom-furniture studio, with clear service positioning and prominent quote requests.

Laureg's Pieces fashion website homepage
WEB 02
livefashion

Laureg’s Pieces

A visual fashion and lifestyle storefront that presents handmade collections clearly and guides customers toward ordering on WhatsApp.

Perfectum Cre8ive studio website homepage
WEB 03
livecreative studio

Perfectum Cre8ive

A bold portfolio and enquiry website for a book-design studio, built to showcase visual work and move prospective clients toward a quote.

03 / What I deliver

Proven starting points, configured around your business.

I begin from tested workflow JSON rather than a blank canvas. Each engagement adapts the rules, prompts, credentials, data and handoff to the client, then ships with documentation and a recorded walkthrough.

01
availablevoice

Voice Receptionist Deployment

Configure the working receptionist for your services, hours, qualification questions, calendar, escalation rules and owner notifications.

Best for: Dental practices · clinics · med spas · home services
Outcome: Fewer missed calls and a direct path from conversation to booked appointment
02
availablelead ops

Lead Response Deployment

Connect the existing capture, classification, reply, owner-alert and tracking workflow to your form, inbox and business rules.

Best for: Ecommerce · agencies · service businesses · teams receiving web enquiries
Outcome: Faster first response and a visible record of every lead
03
availablesupport

Grounded Support Agent Deployment

Turn your approved FAQs and policies into a support agent that uses memory and tool-calling, returns one grounded answer, and refuses unsupported commitments.

Best for: Ecommerce · education · internal support · service businesses
Outcome: Routine questions answered consistently from information your team controls
04 / About

A builder first. I teach because I build, not the other way round.

approach

I build on n8n. It handles the things client work actually runs into — real branching logic, custom code where a no-code node runs out, and no per-task pricing that punishes a system for being used. I teach automation across Zapier, Make and n8n, so if a client needs to own and edit a system inside a tool they already pay for, I can build it there and I say so up front.

how I work

I start from the failure modes, not the happy path. What happens when the model returns nonsense, when the API is down, when someone submits the form twice, when the person who owns the process leaves. A system that only works on a good day is not finished. Every build is handed over with a recorded walkthrough and documentation, because a system nobody understands is a system that quietly dies.

based
Uyo, Nigeria — working globally, UK and US hours
automation
n8n · Zapier · Make
ai
AI agents · multi-agent systems · RAG pipelines · prompt design · OpenAI, Claude, Gemini
voice
Vapi · Retell · ElevenLabs · Twilio
data
Airtable · Notion · Supabase · Pinecone · Google Sheets
digital experiences
Responsive websites · service positioning · conversion paths · client-ready handoff
05 / Teaching

I train other automation specialists.

cybervast · cohort 3

AI Automation Instructor, Cybervast Cohort 3. I teach n8n workflow automation, AI-agent development, and prompt engineering to 13 learners, most from non-technical backgrounds. Ten of thirteen learners have shipped multiple working workflows. I teach across Zapier, Make and n8n; I build on n8n.

06 / Contact

Have a system or digital experience worth building?

Tell me where calls, leads, support questions, or a weak customer journey are slowing the business down. I will show you the closest working solution and what it would take to deploy it.