Syfur RahmanFragmented briefs and review made work harder to coordinate. Knowing those pain points helped us define what to change.
Applied AI
Building with AI.Rethinking work.
I bring experience of marketing operations into hands on AI development. I start with the pain points, question how the work is done and build tools to remove low value tasks and automate repetition.
01 · From operational problem to working platform
Small team. Working platforms.
Working with a colleague, I am using AI to build platforms quickly. I bring the operational knowledge and help shape the workflow, with my colleague leading much of the product delivery. Our shared campaign workflow is already in use.
Bring briefing, review and ownership into one shared workflow. Question which steps still earn their place.
The shared workflow supports campaign delivery today. AI campaign generation is the next capability in development.
Effort modelling points to more capacity. The next test is measured time saved, including checking and corrections.
02 · AI assisted development
See where the effort goes.
I built a workload dashboard with AI to support decisions about capacity, priorities and where automation could help.
Campaign counts show volume. Planning also needs a view of complexity, recorded effort and delivery performance.
A shared view of demand, hours and exceptions, so managers can explore the work behind the totals.
Use the evidence to question recurring tasks, review priorities and choose where to investigate automation.
Work planningWorkload overview
Illustrative dataLast 12 monthsAll groupsAll teamsWork type
Requests2,418+8% month on month
Logged hours7,28682% mapped
Complex work37%analysis · delivery · review
On time98.7%within the agreed target
Demand and effortJan to Aug
Work mixCurrent month
37%complex
ComplexSimple
Capacity by delivery team
Team A78%
Team B91%
03 · Agent workflow design
Move from answers to actions.
I am exploring how agents could carry out repeatable tasks across a workflow. This proposed design sets out the context, permitted actions and checks needed, with a clear point for a person to take over.
Classify the request and retrieve context
Compare options and choose the next action
Working context and approved prior records
Goals, tasks, checks and dependencies
Can it answer directly?
NoYesRun a permitted action and check the result
Checks passed?
YesNoRetry or reviewReturn a checked result or ask for review
Classify user requests, records and events. Retrieve the context needed.
Compare options and choose the next action
Working state and approved prior results
Goals, tasks, checks and dependencies
Can it answer directly?Yes, outputNo, use tools
Check permissions and run an allowed action
SearchAPI callsCodeDatabaseFilesExternal services
Checks passed?Yes, outputNo, retry within limits or ask for review
Return a checked result or ask for review
Available when permittedAvailable tools
SearchAPI callsCodeDatabaseFilesExternal services
04 · Personal experimentation
Family Learning Hub.
A personal prototype that lets me explore how software can respond to individual needs and make progress easier to understand.
Choosing the next learning activity needs context about the child, their school stage and recent progress.
Connect suggested activities with a parent view. Show the next step and keep the parent involved in the decision.
Do the suggestions suit the child and help them learn? That is the next test for this prototype.
Today's path
Four activities, built around progress.
72%weekly goal
MathsFractions · 12 minsStart
ScienceForces · 10 minsNext
ReadingInference · 15 minsLater
Progress at a glance.
6day streak
84%accuracy
3.2hlearning
Suggested next stepRevisit fractions, then move to mixed numbers.
Recommendation based on recent answers and confidence.
05 · Applying the approach beyond marketing
Rethinking restaurant ordering.
I used AI to build an ordering website and explore how the customer journey could connect to the work inside the restaurant.
Menu choices, prices and order notes must stay clear as an order moves from the customer to the team.
The proposed store workflow carries those details into review, with staff accepting orders and confirming collection times.
The next step is to test whether that handover reduces rekeying, clarification calls and order mistakes.
EXAMPLE MENUMenu · Offers · Community
BUILD YOUR ORDERFind it. Add it.
Find it. Add it.
See your total.
YOUR ORDER3 items£28.00
Example store12 min wait
New order ready to review
ORDER 184 · 18:42
Collection request
1 × Vegetable bowlMedium
1 × Seasonal bowlStandard
1 × Flatbread wrapUpgrade
Note, dressing on the side
How I lead this work
Make room for better work.
My goal is to remove low value work and automate repetitive tasks, giving people more time for customers, judgement and improvement. Operational experience helps me find the pain points. It also gives me a reason to challenge habits that no longer serve us.
Work through the detail with the people doing the task. Agree what needs to improve.
Remove work that adds no value before deciding what to automate.
Put a focused tool in front of users early. Use their feedback to shape the next version.
Set permissions and checks. Make exceptions visible and give someone ownership.
Measure adoption, quality and total effort, including review, before expanding.
Marketing leadership in practice