The Old Diocesan Issue 13 - Magazine - Page 53
HOT TOPIC
THE OD NETWORK IN ACTION
Just as Chris and Kieran encourage clients to make connections to boost AI
productivity, so their nascent business is the ideal example of ODs connecting
via their network to boost their chances of success. All the links on this
diagram are a connection made, or still being made, through the OD network.
Kwanda
Cape Town AI consultancy
Chris Immelman
Kieran Donnelly
(2012G) Co-founder
(2010O) Co-founder
Event in April =
effective publicity
The Mitre
Ad-hoc work relationship
ODU HQ,
Cape Town
Potential
collaboration New staff
member
Rafiki Works
London
Brett Chambers
Simon Innes
Greg Cooke
Nick Boswell
(1990S) Winchester
(2013G)
Data engineer
(2011G)
Co-founder
(2010S)
Co-founder
CHRIS AND KIERAN ESTIMATE
that 98% of AI users work only
with generative AI – that is, they
type a prompt into a box, and
receive words or an image in
return. “Digital tennis,” they call
it. You serve; the machine returns.
Unless you hit another ball across
the net, nothing else happens.
The terminology becomes
slippery after that, and Chris is
careful not to pretend that there’s
one perfect taxonomy. The useful
distinction is between AI that
generates and AI that acts. Give
the brain tools – access to meeting
notes, email, calendars, databases
– and it gains arms. It can convert
a recorded meeting into an
action list. It can read an email in
conjunction with that action list,
prepare your response and place
it in Drafts. It can identify overdue
invoices and draft follow-ups.
Agentic AI goes further,
initiating actions or working
through a task in a loop, reviewing
and improving its own output.
But for most businesses, the
immediate leap is from talking to
AI to allowing it, within carefully
defined permissions, to use tools.
“When an agent can query
data, run tasks across tools
and has context, you unlock
magic,” Chris explains. Get that
working, and you’re already
operating beyond the vast
majority of users.
This is the territory Kwanda
occupies. A typical engagement
with a client begins with up to
a month of discovery: conducting
deep dives with key employees;
getting hands dirty to understand
the key business objectives;
digging down into what a specific
business unit is optimising for.
“This culminates in a roadmap,
which contains a set of initiatives
that map directly back to the
business objectives uncovered
in the discovery,” says Chris.
Where an initiative doesn’t create
business value, it doesn’t end up
in the roadmap.
The ideal result is a 90-day
process, in which Kwanda then
manages the implementation of
a central AI tool, connects it to
the company’s working systems,
builds “skills” or automations, and
trains staff and executives. Larger
clients may need a senior data
engineer, supported by a more
junior engineer, embedded in the
business to reorganise data before
useful intelligence can be built
on top of it. These are “Kwanda
solutions engineers”, a variation
on the increasingly sought-after
“forward-deployed engineer” –
someone technically excellent who
sits close enough to a business to
understand how it really works.
Six months after launch, Kwanda
is already profitable and the two
founders have become a team of
eight, with more joining. Chris, the
only non-engineer, explains that
recent graduates are ideal. Bright
and engaged, and working in an
exciting new sector in which “work
experience” is a relative term, they
are hired “on slope and aptitude”.
* * * * *
SO WHAT SHOULD AN INDIVIDUAL
or a small business do right now?
First, stop assuming everyone
else has disappeared over the
horizon, and actively develop
your AI policy. “You’re not as far
behind as the internet will have
you believe,” Chris says. “And
besides, every company has
an AI policy already, whether
it knows it or not. For most of
them right now, the policy is staff
using the free version of ChatGPT.”
THE OLD DIOCESAN | 49
