AI Implementation Support
Where you have already chosen a platform, or are about to, we act on your side of the table. Vendors demonstrate on their data, not yours, and the gap between a polished demonstration and production behaviour is where budgets disappear. Independent evaluation, acceptance testing against what was actually claimed, and adoption support are what close that gap.
1
Requirements
What you actually need
2
Evaluate
Test the claims
3
Contract
Terms that protect you
4
Accept
Verify before signing off
5
Adopt
Make it used
How We Support You
01
Requirements Definition
We help you write requirements against your process rather than against a vendor feature list. Requirements derived from a demonstration describe that product, which makes genuine comparison impossible.
- Requirements written from your process, not a brochure
- Mandatory separated from desirable
- Data residency and POPIA requirements stated explicitly
- Evaluation criteria agreed and weighted in advance
02
Vendor Evaluation
We assess vendors against those criteria and press on the claims that matter. Accuracy figures quoted without a stated test set mean nothing, and asking for the methodology behind a number is remarkably clarifying.
- Structured evaluation against weighted criteria
- Accuracy claims tested against your own sample data
- Reference checks with comparable clients
- Total cost including integration, training and running
03
Contracting Support
We help you negotiate terms that protect you: data ownership, exit provisions, performance commitments and what happens when the model underperforms. Standard vendor terms rarely address any of those.
- Data ownership and permitted use of your data
- Exit and data extraction provisions
- Performance commitments with defined remedies
- POPIA operator clauses where personal information is processed
04
Acceptance Testing
We test what was delivered against what was claimed, using your data, before you sign acceptance. It is the last moment at which you have genuine commercial leverage.
- Testing against your data rather than vendor samples
- Verification of each claimed capability
- Integration tested end to end
- Defects logged and resolved before acceptance
05
Adoption and Handover
Technology nobody uses delivers nothing. We train by role, measure whether usage actually happens, and review after implementation against the benefit the business case promised.
- Role based training on the delivered system
- Adoption measured from system usage data
- Governance and approval integrated into daily use
- Post implementation review against the business case
What You Receive
- Requirements specification written from your process
- Weighted vendor evaluation with a documented recommendation
- Contract review covering data, exit and performance terms
- Acceptance test results against claimed capability
- Integration delivered and tested end to end
- Post implementation review against the business case
Indicative Timeline
Evaluation and selection normally take four to six weeks. Implementation support runs alongside the vendor project and its duration is set by that project rather than by us.
- Requirements definition: one to two weeks
- Vendor evaluation and reference checks: two to three weeks
- Contracting support: dependent on negotiation
- Acceptance testing and adoption: through to go live
Where We Add Value
Implementation support exists to keep the balance of information even between you and the vendor.
Independent Evaluation
We have no reseller relationship with AI vendors, so the recommendation follows the evaluation.
Claim Verification
Accuracy figures tested against your own data rather than accepted from a demonstration.
Contract Terms
Data ownership, exit rights and performance remedies, which standard vendor terms rarely cover.
Acceptance Testing
Verification before sign off, while you still hold commercial leverage.
Integration
Connecting the platform to the systems it has to work with, which is routinely underestimated.
Adoption
Training and measurement, because unused software is the most common form of wasted spend.
Frequently Asked Questions
Do you resell any AI platforms?
No. We hold no reseller or commission relationship with AI vendors, which is precisely why organisations ask us to sit on their side of the evaluation. Our recommendation follows the criteria rather than a margin.
We have already chosen a vendor. Is it too late?
No. Acceptance testing and integration support are the highest value points, and they come after selection. Being able to demonstrate that delivered capability differs from what was claimed is worth considerably more before sign off than after.
How do we test a vendor accuracy claim?
With your own data. Ask for a proof of concept on a sample you provide, with the test set and methodology stated. Vendors confident in their product agree readily; the reluctance itself is informative.
What should we insist on in the contract?
Ownership of your data and of anything derived from it, the right to extract it on exit in a usable format, performance commitments with remedies, and POPIA operator provisions where personal information is processed.
What if the platform underperforms after go live?
That depends entirely on what the contract says, which is why the contracting stage matters. Without defined performance commitments and remedies, your only options are renegotiation or replacement, both expensive.
Can you manage the whole implementation?
We can run the project management and integration, working with the vendor. Where the platform is genuinely turnkey, lighter oversight is more appropriate and we will say so rather than manufacture a role.
Related Services
This sits inside our AI Services practice. Related work: AI Readiness Assessment to establish whether to buy at all, and AI Governance and Policy for the framework a new platform has to operate within.
