Department AI Training
Built around one function real work, not general training with their logo on it.
What is department training?
Department AI training is built around the specific tasks of one function, recognising that the useful applications and the acceptable risks differ substantially between finance, marketing, HR, legal and operations.
General AI training treats every department as though the work were the same. It is not. Marketing benefits from drafting and variation; finance benefits from summarising and checking; legal benefits from almost none of the same things and carries different risk.
The differences that matter most are about tolerance for error. A marketing draft that is wrong gets edited. A figure in a management account that is wrong gets acted on.
Training that ignores this teaches everybody the same habits, which are appropriate for some functions and actively unsafe for others.
Why no two departments get the same session
Duara trains individual functions from Nairobi, most often finance, marketing, HR, customer operations and legal teams.
We refuse to reuse one deck across departments. The reason it is common is that it is cheap to deliver, and the result is finance staff taught the checking habits appropriate to a marketing draft.
Who this is for
- One function needs practical AI use
- General training did not fit the work
- Different departments face different risks
- You want to start with one team rather than everybody
Industries we serve
- Financial services
- Manufacturing
- Professional services
- Healthcare administration
- Retail groups
- NGOs and development
Problems this solves
These describe training pitched at an organisation when the work happens in functions.
- Generic examples that do not land. Staff are shown a demonstration on a task nobody in the room actually does.
- Uniform habits across different risks. The same checking approach is taught to marketing and to finance, where the consequence of a wrong output differs entirely.
- Applications nobody identified. The genuinely useful use for that function was never worked out, so people default to obvious and low value ones.
- One team ready and the rest not. A single function is keen and organisation wide rollout is not justified yet.
What we deliver
Training built on the function own tasks, checking habits matched to the consequence of error in that work, applications identified with the team, and a starting point that does not require organisation wide commitment.
Training built on that function own tasks
Checking habits matched to the work risk
Useful applications identified with the team
Function specific guidance on what not to enter
A reference sheet using their own examples
A basis for deciding whether to extend further
Features
- Built per function, not reused. One deck across departments is cheap to deliver and teaches finance the habits appropriate to a marketing draft.
- Checking matched to consequence. A wrong marketing draft gets edited. A wrong figure in a management account gets acted on, and the habits should differ accordingly.
- Applications found with the team. The people doing the work identify the useful cases faster than any external assessment.
- Function specific boundaries. What HR must not enter differs from what marketing must not, and generic rules cover neither well.
- A sensible starting unit. One function is enough to learn from without committing the organisation to anything.
Benefits
- Examples people recognise. Training on the work they actually do transfers in a way a generic demonstration does not.
- Safer habits where it matters. Functions handling figures or personal data get the checking discipline their work requires.
- Better applications found. The team identifies uses an external trainer would not have thought of.
- A low commitment start. One department tells you whether wider rollout is worth funding.
Technologies we use
- Task mapping per function. Identifying the specific work where AI genuinely helps that team.
- Risk calibration. Setting checking expectations against the consequence of error in that function.
- Function boundaries. What that team specifically must not enter, which differs by the data they handle.
- Applied practice. Working on that function real tasks rather than general examples.
- Extension assessment. Judging from one function whether wider rollout is justified.
The variable that changes between functions is not the tool, it is the cost of being wrong. Training that does not adjust for that is the same training in a different room.
How the process works, step by step
Department training is usually half a day per function, with the task mapping done beforehand rather than in the room.
-
Step 1
Free function review
We establish what that team actually does daily and where the errors would be costly.
-
Step 2
Tasks mapped with the team
Useful applications are identified with the people doing the work rather than assumed.
-
Step 3
Risk calibrated
Checking expectations are set against what a wrong output would cost in that function.
-
Step 4
Session delivered
Practice happens on their own tasks, with their own examples and their own boundaries.
-
Step 5
Extension decided
You decide from one function result whether wider rollout is worth funding.
Quality assurance and how we work
The commitments attached to training one function properly.
- We build per function rather than reuse. A single deck across departments is cheap to deliver and produces checking habits that are wrong for half the audience.
- Checking expectations follow the risk. The consequence of a wrong output differs enormously between functions and the training has to reflect that.
- We identify applications with the team. People doing the work find the useful cases faster than an external assessment does.
- We do not push organisation wide rollout. One function is a legitimate endpoint, and we would rather it worked than sold you the next four.
- We say when a function gains little. Some work benefits genuinely less, and pretending otherwise wastes a session and damages credibility.
Frequently asked questions about department training
Why not train everybody the same way?
Because the work differs. Marketing benefits from drafting and variation, finance from summarising and checking, and the risks are not comparable.
What actually changes between functions?
The cost of being wrong. A marketing draft that is wrong gets edited; a figure in a management account that is wrong gets acted on.
Is one deck across departments really a problem?
Yes. It is cheap to deliver and it teaches finance the checking habits appropriate to a marketing draft.
Which functions benefit most?
Ones producing repeated written work or summarising volume. Marketing, HR and customer operations usually show the fastest return.
Which benefit least?
Highly precision critical work where checking costs as much as doing. We say so rather than running a session that disappoints.
How long per function?
Half a day, with the task mapping done beforehand rather than consuming session time.
Who should be in the room?
The people doing the work rather than only the managers, since the useful applications come from them.
Do we need to train everybody afterwards?
No. One function is a legitimate endpoint, and we would rather it worked than sell you the next four.
Can it cover our function specific rules?
Yes, and it should. What HR must not enter differs from what marketing must not, and generic rules cover neither well.
What if the team is sceptical?
Common, and working on their real tasks addresses it faster than argument. Scepticism about generic examples is usually justified.
Can several departments be trained in sequence?
Yes, and each gets built separately. Sequencing them is fine; reusing the content is not.
Does finance need different handling?
Yes. Figures acted on rather than edited means checking discipline matters far more, and that shapes the whole session.
What about legal or compliance teams?
They benefit narrowly and carry higher risk, so the boundaries take up more of the session than the applications do.
Will you find uses we have not thought of?
Sometimes, though more often the team finds them once given the framing. That is a better outcome anyway.
Do you need access to our systems?
No. Training works on documents and tasks the team brings rather than requiring system access.
Can it be run for a small team?
Yes, and small teams work well since everybody gets attention on their own tasks.
How does this differ from your organisation wide training?
That handles the range of ability at scale. This handles the specific work and risk of one function.
Will the team retain it?
Better than generic training, since the examples are their own. Practice afterwards still matters.
Is what the team says confidential?
Patterns are reported without naming individuals, and complaints raised in the room stay in it.
What does department training cost?
Function count and group size determine it, quoted after the function review.
Glossary of terms
Terms used in function specific training.
- Function
- A department or team defined by the work it does, such as finance or marketing.
- Risk calibration
- Matching how carefully output is checked to what being wrong would cost.
- Task mapping
- Identifying the specific pieces of work where a tool genuinely helps a team.
- Boundary
- What a particular function must not enter into a tool, given the data it handles.
- Applied practice
- Working on real tasks from that function rather than general examples.
- Extension
- Deciding whether to roll further after one function has been trained.
Related services in AI Training
Talk to us about department training
Pick the function where the same document gets written every week. That is usually where department training pays back fastest.
Start your project