{
}
_
</>

Alfie Roberts — AI Workflow Consultant for PR and Communications Agencies

ai workflow consultancy for pr and communications agencies

const
async
import
=> {
return
AI
ML
()
Alfie Roberts

About Alfie Roberts — AI Workflow Consultant

i work with pr and communications agencies to automate the workflows that eat their teams' time and help them figure out how to effectively use ai to speed things up. from coverage tracking, measurement reporting, media intelligence, to insight generation, i can help with every stage of the process. the result is less manual admin, better data, and more time for the work that actually requires a human.

previously ai and advanced analytics lead at ketchum, part of the omnicom group, where i built measurement systems and ai tools for brands including p&g, samsung, and danone. i've since taken that experience independent, working directly with agencies on the automation and data problems that off-the-shelf tools don't solve.

i'm experienced in python, machine learning, traditional ai and logic, as well as azure, and data pipelines. my current focus is applied ai, specifically large language models, agentic workflows, and automation in agency contexts.

outside of work i'm either tinkering with my homelab or being bossed around by amalfi the cat.

[

AI Workflow Automation Services

]

i help agencies figure out which parts of their workflow can be automated, build the tools to do it, and make sure those tools actually get used.

that typically means: measurement and reporting systems, media intelligence pipelines, ai-assisted research and insight tools, and coverage tracking infrastructure. i work across the full stack — from scoping the problem to handing over something that runs.

ai workflow consultancy is the practice of identifying which parts of an agency's daily operations — reporting, tracking, data collection, insight generation — can be partially or fully automated using ai and software. unlike off-the-shelf saas tools, a consultant builds systems tailored to how a specific team actually works, connecting existing platforms like google workspace, office 365, and media monitoring tools into automated pipelines. at ketchum i built these systems for brands like p&g, samsung, and danone — cutting monthly reporting from days to hours, winning an amec gold award for best use of new technology, and proving that the right automation gives teams more time for creative and strategic work, not less.

pythonsqlpysparkmlnlpazurepowerbidata storytellingstrategyresearchinsightsconsulting
{

Frequently Asked Questions About AI Workflow Automation

}

does ai automation need a huge budget?

not usually. the quickest wins come from tools you're already paying for. google workspace and office 365 both have powerful automation built in — most teams just don't know it's there. the real work is documenting your processes and translating them into automations inside the platforms you already use. the expensive stuff comes later, if it's needed at all.

what does an ai workflow consultant actually do?

i look at how your team works day-to-day — the reporting, the tracking, the manual admin — and figure out which parts can be automated. then i build it, hand it over, and make sure it actually gets used. that might mean a measurement dashboard, a media monitoring pipeline, or just getting your existing tools to talk to each other.

will this replace people on my team?

no, and i won't work on projects that try to. i don't automate creativity, thinking, or judgment. the goal is to remove the repetitive admin that eats your team's time so they can focus on the work that actually needs a human. if a brief asks me to replace people or enable teams to stop thinking, i'll turn it down.

what do i need to have in place before we start?

nothing. i prefer starting from scratch — understanding your team, your processes, and your goals before building anything. i'm happy to pick up unfinished work or improve what's already there, but the ideal starting point is a team with ideas they don't know how to realise yet.

how long does an engagement take?

it starts with a 30-minute scoping call. after that, it moves as fast as you need — some clients run workshops the following week, others prefer a longer timeline. pricing is day rate for consultancy and project-based for development, both discussed on the call.

AI and Data Engineering Projects

{ }

danone measurement platform

built a measurement platform that tracks how danone's brands perform across pr and influencer campaigns. won an amec gold award for "best use of new technology" — the oscars of the measurement world.

pythonanalyticsaward winnerview →
{ }

train delay predictor

can we predict when trains will be delayed before it happens? turns out, yes — this ml model catches 76% of incidents before they're announced. built as a fully deployable api.

mlflaskdockerview →
{ }

supplies for ukraine c.i.c.

when the invasion started, i co-founded a platform to get medical supplies to ukraine. taught myself javascript to build it. £80k+ raised and counting.

javascriptnode.jsimpactview →
{ }

kr sensing

an early-warning system for p&g beauty that monitors social, news, academic, and policy sources in real-time. data flows through a bronze→silver→gold pipeline in azure synapse, where seven composite indicators feed into a 0-100 risk score with anomaly detection. if something's brewing, this catches it first. gave the brand team weeks of early warning on emerging ingredient and regulatory risks that would previously have arrived as surprises.

azure synapseaireal-time analytics
{ }

samsung earned media tracker

built the measurement backbone for samsung's uk pr team. automated data collection from 15+ sources, standardised reporting across product launches, and created executive dashboards that actually got used. cut the team's monthly reporting time from days to hours and gave senior stakeholders a single view of earned media performance across all product lines.

azurepysparkpowerbi
{ }

oracle ad spend forecaster

regression models that predicted quarterly ad spend trends across oracle's client base. the forecasts directly shaped sales strategy — and consistently beat the gut-feel estimates.

pythonregressionforecasting
{ }

json to pyspark schema generator

a python tool that generates pyspark schemas from complex nested json data. handles arrays, nested objects, and outputs ready-to-use dataframe schemas. built on top of preetranjan's javascript implementation.

pythonpysparkopen sourceview →
[

PR and Communications Agency Clients

]
( )

Contact Alfie Roberts — AI Workflow Consultancy

got a project in mind? i'd love to chat.

book a chat