A quantitative survey of 415 professionals

The study asks how artificial intelligence tools are adopted and integrated across the stages of the marketing process framework in South Africa’s e-commerce sector. It uses the SOSTAC framework (Situational Analysis, Objectives, Strategy, Tactics, Action, Control) and a self-administered online survey of 415 digital marketing professionals working in South African retail, using a four-point Likert scale (from 1, strongly disagree, to 4, strongly agree). The protocol received university ethics approval (NWU-00620-24-A4). [1]

The sample was drawn through non-probability purposive sampling from the online panel of an international market-research company, with respondents required to hold a role involving digital marketing in South African retail. [1]

Where AI is most used

Fourteen AI applications were each rated on the 1-4 scale. The highest reported agreement levels (combining 'agree' and 'strongly agree') were for content personalisation (88.2%) and content marketing for a targeted audience (87.7%). [1]

Mapped onto the framework’s six stages, this adoption is heavily concentrated in the Tactics and Action stages (content personalisation, chatbots), which the study describes as showing high adoption. [1]

A classification gap at the Objectives stage

The AI applications table does not link any of the fourteen surveyed applications to the Objectives stage of the SOSTAC framework; the author reads this as a potential gap in AI’s use for goal-setting, not as a measured, lower adoption level for this stage: the study reports no adoption data for Objectives. [1]

Voice search and emails: the least favoured tactical applications

Among the fourteen applications surveyed, voice search and personalised emails — classified by the study under the framework’s Tactics stage, alongside content personalisation — show the lowest reported agreement levels, with means of 3.02 and 3.01 respectively on a 1-4 scale. [1]

The study states that these two applications show more variability in their perceived effectiveness or ease of implementation, which it links to lower mean values and higher standard deviations. [1]

The descriptive-statistics table qualifies this reading: voice search’s standard deviation (0.834) is lower than that of customer-support chatbots (0.846), an application the study otherwise presents as a pillar of high adoption; among the fourteen applications, only personalised emails (0.860) and chatbots (0.846) have the two highest standard deviations. [1]

Strategy and control: the author’s qualitative summary

In the author’s qualitative summary by stage, Strategy (predictive analytics for segmentation) is described as ‘moderate/emerging’ adoption. Situational analysis (understanding consumer behaviour patterns) is described as ‘moderate’. Control (real-time optimisation and attribution) is described as ‘emerging/lower’. [1]

What the author’s own recommendations add

The author recommends that organisations prioritise investment in data infrastructure, cross-functional integration and AI literacy to enable more strategic uses of AI, and embed AI systematically across the planning, execution and evaluation stages, within regulatory frameworks that balance personalisation with privacy. [1]

Figures the study’s own text does not always confirm

The study’s text states that content personalisation (AIA2) has the highest mean, at 3.17 out of 4. Yet the same study’s descriptive-statistics table reports a higher mean, 3.19 out of 4, for customer support via chatbots (AIA4): the two published figures do not match. [1]

Similarly, the text identifies understanding consumer behaviour (AIA14) as the application with the highest share of 'strongly disagree' responses, at 6.3%, while the study’s own table shows a higher share, 6.7%, for personalised emails (AIA7). [1]

The study reports only descriptive statistics (percentages, means, standard deviations, skewness, kurtosis) on self-reported agreement; no statistical test, correlation or comparison between stages is presented, which limits any reading in terms of a tested pattern or progression between stages. [1]

Limits of the survey

The non-probability purposive sampling, drawn from a single research company’s panel and restricted to South African retail respondents, means the results describe this panel’s perceptions and cannot be generalised to all digital marketing professionals, other sectors, or other countries. [1]

The reported adoption levels rest on self-reported agreement statements ('our organisation applies...') rather than verified usage, spend or performance data: the 86-88% figures reflect individual perception, not audited deployment. [1]

What a planning team can take from this

Before extending AI to control and attribution functions, a team running its marketing plan on the SOSTAC framework could explicitly check, at its own Objectives stage, whether an AI tool actually supports goal-setting, rather than assuming AI coverage is uniform across the framework: this is not a finding of the study but an internal check suggested by the editorial team. [1]