26
AI &AUTOMATION BAROMETER REPORT
• Don’t see how the bulk of it can be best applied to our business without lessening quality and service.
• Finetuning and control, among other issues.
• Hype, results often disappoint. Needs lots of finetuning.
• Implementation is far more complex than AI sellers divulge
• Lack of compatibility in the tech-stack
• Low-quality AI for our languages
• Mastering the ability to interact with such systems to streamline processes rather than creating
rabbit holes. Determine the use cases eligible for AI workflows. Share with people not directly
involved in the AI workflows good and bad requests based on the approved/disapproved use cases.
• Quality. Disambiguation of the marketing text of the services (Does it work well, and how much
adaption is needed to adopt it properly).
• Tests, basic integration and functional analysis
• That you can not trust the output and the behaviour.
• We have been using the already developed integration channels but many times the defined
document workflow doesn’t work or doesn’t provide the output as stated.
Lack of resources
• Costs (x3)
• Demand for developers
• Development resources
• Finding people with the right mindset, knowledge and skills
• How to bill the work when AI is used (x2)
• Lack of infrastructure and resources, shortage of skilled personnel, delayed or extended launches,
resource constraints.
• Lack of resources
• Planning and budgeting.
Data &security concerns
• Client safety concerns
• Collection and development of training data.
• Finding a third-party solution that can be trained on own datasets, and be integrated in current tools.
• Inbuilt AI in the CAT tool actually sends data to the open servers, this is why we are rarely using this
function. Only if the files /content is not confidential
Staff learning curve
• Changing habits amongst employees. Implementing new processes.
• General scepticism to the negative effects
CHALLENGES
AI &AUTOMATION BAROMETER REPORT
• Don’t see how the bulk of it can be best applied to our business without lessening quality and service.
• Finetuning and control, among other issues.
• Hype, results often disappoint. Needs lots of finetuning.
• Implementation is far more complex than AI sellers divulge
• Lack of compatibility in the tech-stack
• Low-quality AI for our languages
• Mastering the ability to interact with such systems to streamline processes rather than creating
rabbit holes. Determine the use cases eligible for AI workflows. Share with people not directly
involved in the AI workflows good and bad requests based on the approved/disapproved use cases.
• Quality. Disambiguation of the marketing text of the services (Does it work well, and how much
adaption is needed to adopt it properly).
• Tests, basic integration and functional analysis
• That you can not trust the output and the behaviour.
• We have been using the already developed integration channels but many times the defined
document workflow doesn’t work or doesn’t provide the output as stated.
Lack of resources
• Costs (x3)
• Demand for developers
• Development resources
• Finding people with the right mindset, knowledge and skills
• How to bill the work when AI is used (x2)
• Lack of infrastructure and resources, shortage of skilled personnel, delayed or extended launches,
resource constraints.
• Lack of resources
• Planning and budgeting.
Data &security concerns
• Client safety concerns
• Collection and development of training data.
• Finding a third-party solution that can be trained on own datasets, and be integrated in current tools.
• Inbuilt AI in the CAT tool actually sends data to the open servers, this is why we are rarely using this
function. Only if the files /content is not confidential
Staff learning curve
• Changing habits amongst employees. Implementing new processes.
• General scepticism to the negative effects
CHALLENGES
